Remove --skill flag from skill subcommand to install all skills

The skill command now runs 'npx skills add mark3labs/kit' without
filtering to a single skill, installing both available skills:

1. Extensions - creating Kit extensions
2. SDK - building with the Kit Go SDK

.gitignore

@@ -1,7 +1,8 @@
 .aider*
 .task/
 .env
-.kit/
+.kit/*
+!.kit/extensions/
 aidocs/
 *.log
 /kit

.kit/extensions/go-edit-lint.go

@@ -0,0 +1,228 @@
+//go:build ignore
+
+package main
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"os/exec"
+	"path/filepath"
+	"strings"
+	"time"
+
+	"kit/ext"
+)
+
+const (
+	diagnosticsTimeout = 20 * time.Second
+	maxOutputBytes     = 12_000
+)
+
+type toolPathInput struct {
+	Path string `json:"path"`
+}
+
+type lintResult struct {
+	Output string
+	Err    error
+}
+
+func Init(api ext.API) {
+	api.OnSessionStart(func(_ ext.SessionStartEvent, ctx ext.Context) {
+		ctx.Print("go-edit-lint extension loaded - will run gopls and golangci-lint on Go file edits")
+	})
+
+	api.OnToolResult(func(e ext.ToolResultEvent, ctx ext.Context) *ext.ToolResultResult {
+		if e.IsError || !isEditOrWrite(e.ToolName) {
+			return nil
+		}
+
+		absPath, ok := resolveGoFilePath(e.Input, ctx.CWD)
+		if !ok {
+			return nil
+		}
+
+		report := runGoDiagnostics(ctx.CWD, absPath)
+		
+		// Check if there are issues and add explicit prompt for the LLM to react
+		goplsIssues, lintIssues := countIssues(report)
+		hasIssues := goplsIssues > 0 || lintIssues > 0
+		
+		var enhanced string
+		if hasIssues {
+			enhanced = e.Content + "\n\n" + report + "\n\n⚠️ DIAGNOSTICS FOUND: Please review the issues above and fix them before proceeding."
+		} else {
+			enhanced = e.Content + "\n\n" + report
+		}
+
+		// Show TUI message block for diagnostics visibility (only if there are issues)
+		if hasIssues {
+			var msgLines []string
+			msgLines = append(msgLines, fmt.Sprintf("File: %s", filepath.Base(absPath)))
+			if goplsIssues > 0 {
+				msgLines = append(msgLines, fmt.Sprintf("gopls: %d issue(s)", goplsIssues))
+			}
+			if lintIssues > 0 {
+				msgLines = append(msgLines, fmt.Sprintf("golangci-lint: %d issue(s)", lintIssues))
+			}
+			msgLines = append(msgLines, "", "⚠️ Please fix these issues before proceeding.")
+
+			borderColor := "#f9e2af" // yellow
+			if goplsIssues > 0 && lintIssues > 0 {
+				borderColor = "#f38ba8" // red
+			}
+
+			ctx.PrintBlock(ext.PrintBlockOpts{
+				Text:        strings.Join(msgLines, "\n"),
+				BorderColor: borderColor,
+				Subtitle:    "go-edit-lint",
+			})
+		}
+
+		return &ext.ToolResultResult{Content: &enhanced}
+	})
+}
+
+func isEditOrWrite(toolName string) bool {
+	return strings.EqualFold(toolName, "edit") || strings.EqualFold(toolName, "write")
+}
+
+func resolveGoFilePath(inputJSON, cwd string) (string, bool) {
+	var args toolPathInput
+	if err := json.Unmarshal([]byte(inputJSON), &args); err != nil || args.Path == "" {
+		return "", false
+	}
+
+	absPath := args.Path
+	if !filepath.IsAbs(absPath) {
+		absPath = filepath.Join(cwd, absPath)
+	}
+
+	if strings.ToLower(filepath.Ext(absPath)) != ".go" {
+		return "", false
+	}
+
+	return absPath, true
+}
+
+func runGoDiagnostics(cwd, absPath string) string {
+	gopls := runGopls(cwd, absPath)
+	lint := runGolangCILint(cwd, "./...")
+
+	return fmt.Sprintf(
+		"<go_diagnostics file=%q>\n[gopls]\n%s\n\n[golangci-lint]\n%s\n</go_diagnostics>",
+		filepath.Base(absPath),
+		formatToolResult(gopls, "No diagnostics."),
+		formatToolResult(lint, "No lint issues."),
+	)
+}
+
+func runGopls(cwd, absPath string) lintResult {
+	ctx, cancel := context.WithTimeout(context.Background(), diagnosticsTimeout)
+	defer cancel()
+
+	cmd := exec.CommandContext(ctx, "gopls", "check", absPath)
+	cmd.Dir = cwd
+	out, err := cmd.CombinedOutput()
+
+	if ctx.Err() == context.DeadlineExceeded {
+		return lintResult{Err: fmt.Errorf("timed out after %s", diagnosticsTimeout)}
+	}
+
+	if err != nil {
+		return lintResult{Output: truncate(string(out), maxOutputBytes), Err: fmt.Errorf("failed to run gopls check: %w", err)}
+	}
+
+	return lintResult{Output: truncate(string(out), maxOutputBytes)}
+}
+
+func runGolangCILint(cwd, target string) lintResult {
+	ctx, cancel := context.WithTimeout(context.Background(), diagnosticsTimeout)
+	defer cancel()
+
+	args := []string{
+		"run",
+		target,
+		"--show-stats=false",
+		"--output.text.path", "stdout",
+		"--output.text.colors=false",
+		"--output.text.print-issued-lines=false",
+	}
+	cmd := exec.CommandContext(ctx, "golangci-lint", args...)
+	cmd.Dir = cwd
+	out, err := cmd.CombinedOutput()
+
+	if ctx.Err() == context.DeadlineExceeded {
+		return lintResult{Err: fmt.Errorf("timed out after %s", diagnosticsTimeout)}
+	}
+
+	trimmed := truncate(string(out), maxOutputBytes)
+	if err == nil {
+		return lintResult{Output: trimmed}
+	}
+
+	exitErr, ok := err.(*exec.ExitError)
+	if ok && exitErr.ExitCode() == 1 {
+		return lintResult{Output: trimmed}
+	}
+
+	return lintResult{Output: trimmed, Err: fmt.Errorf("failed to run golangci-lint: %w", err)}
+}
+
+func formatToolResult(res lintResult, emptyFallback string) string {
+	var lines []string
+	if res.Err != nil {
+		lines = append(lines, "ERROR: "+res.Err.Error())
+	}
+	out := strings.TrimSpace(res.Output)
+	if out == "" {
+		if res.Err == nil {
+			lines = append(lines, emptyFallback)
+		}
+	} else {
+		lines = append(lines, out)
+	}
+	if len(lines) == 0 {
+		return emptyFallback
+	}
+	return strings.Join(lines, "\n")
+}
+
+func truncate(s string, max int) string {
+	if len(s) <= max {
+		return s
+	}
+	return s[:max] + "\n... output truncated ..."
+}
+
+func countIssues(report string) (goplsCount, lintCount int) {
+	// Extract gopls section
+	goplsStart := strings.Index(report, "[gopls]")
+	lintStart := strings.Index(report, "[golangci-lint]")
+	endTag := strings.Index(report, "</go_diagnostics>")
+
+	if goplsStart != -1 && lintStart != -1 {
+		goplsSection := report[goplsStart:lintStart]
+		// Count non-empty lines excluding the header and "No diagnostics." message
+		for _, line := range strings.Split(goplsSection, "\n") {
+			line = strings.TrimSpace(line)
+			if line != "" && line != "[gopls]" && line != "No diagnostics." {
+				goplsCount++
+			}
+		}
+	}
+
+	if lintStart != -1 && endTag != -1 {
+		lintSection := report[lintStart:endTag]
+		// Count non-empty lines excluding the header and "No lint issues." message
+		for _, line := range strings.Split(lintSection, "\n") {
+			line = strings.TrimSpace(line)
+			if line != "" && line != "[golangci-lint]" && line != "No lint issues." {
+				lintCount++
+			}
+		}
+	}
+
+	return goplsCount, lintCount
+}

README.md

@@ -21,7 +21,9 @@ A powerful, extensible AI coding agent CLI with multi-provider support, built-in
 - **Built-in Core Tools**: bash, read, write, edit, grep, find, ls, spawn_subagent - no MCP overhead
 - **MCP Integration**: Connect external MCP servers for expanded capabilities
 - **Extension System**: Write custom tools, commands, widgets, and UI modifications in Go
-- **Theming**: 22 built-in color themes (KITT, Catppuccin, Dracula, Nord, etc.) with runtime switching and custom theme files
+- **Theming**: 22 built-in color themes (KITT, Catppuccin, Dracula, Nord, etc.) with runtime switching, persistence, and custom theme files
+- **Model Persistence**: Model and thinking level selections are automatically saved and restored across sessions
+- **Prompt Templates**: Create reusable prompt templates with shell-style argument substitution
 - **Interactive TUI**: Rich terminal interface powered by Bubble Tea with streaming, syntax highlighting, and custom rendering
 - **Session Management**: Tree-based conversation history with branching support
 - **Non-Interactive Mode**: Script-friendly positional args with JSON output
@@ -71,8 +73,11 @@ kit @main.go @test.go "Review these files"
 # Continue the most recent session
 kit --continue
 
+# Model and thinking level selections are automatically persisted
+# across sessions and restored on next launch
+
 # Use specific model
-kit --model anthropic/claude-sonnet-4-5-20250929
+kit --model anthropic/claude-sonnet-latest
 ```
 
 ### Non-Interactive Mode
@@ -116,7 +121,7 @@ Kit looks for configuration in the following locations (in order of priority):
 Create `~/.kit.yml`:
 
 ```yaml
-model: anthropic/claude-sonnet-4-5-20250929
+model: anthropic/claude-sonnet-latest
 max-tokens: 4096
 temperature: 0.7
 stream: true
@@ -177,6 +182,8 @@ mcpServers:
 # Extensions
 --extension, -e          Load additional extension file(s) (repeatable)
 --no-extensions          Disable all extensions
+--prompt-template        Load a specific prompt template by name
+--no-prompt-templates    Disable prompt template loading
 
 # Generation parameters
 --max-tokens             Maximum tokens in response (default: 4096)
@@ -232,6 +239,8 @@ Kit ships with 22 built-in color themes that control all UI elements. Switch at
 /theme tokyonight
 ```
 
+Theme selections are automatically saved and restored on next launch (stored in `~/.config/kit/preferences.yml`). This persistence also applies to **model** and **thinking level** selections — all are saved together and restored on startup.
+
 ### Custom themes
 
 Drop a `.yml` file in `~/.config/kit/themes/` (user) or `.kit/themes/` (project):
@@ -278,7 +287,7 @@ kit -e examples/extensions/minimal.go
 
 ### Extension Capabilities
 
-**Lifecycle Events**: OnSessionStart, OnSessionShutdown, OnBeforeAgentStart, OnAgentStart, OnAgentEnd, OnToolCall, OnToolExecutionStart, OnToolExecutionEnd, OnToolResult, OnInput, OnMessageStart, OnMessageUpdate, OnMessageEnd, OnModelChange, OnContextPrepare, OnBeforeFork, OnBeforeSessionSwitch, OnBeforeCompact
+**Lifecycle Events**: OnSessionStart, OnSessionShutdown, OnBeforeAgentStart, OnAgentStart, OnAgentEnd, OnToolCall, OnToolExecutionStart, OnToolOutput, OnToolExecutionEnd, OnToolResult, OnInput, OnMessageStart, OnMessageUpdate, OnMessageEnd, OnModelChange, OnContextPrepare, OnBeforeFork, OnBeforeSessionSwitch, OnBeforeCompact, OnCustomEvent
 
 **Custom Components**:
 - **Tools**: Add new tools the LLM can invoke
@@ -332,6 +341,8 @@ See the `examples/extensions/` directory:
 - `tool-renderer-demo.go` - Custom tool call rendering
 - `widget-status.go` - Persistent status widgets
 
+Also see `.kit/extensions/go-edit-lint.go` (in this repo) for a project-local extension example that runs gopls and golangci-lint on Go file edits.
+
 ### Loading Extensions
 
 **Auto-discovery** (loads automatically):
@@ -389,6 +400,32 @@ func TestMyExtension(t *testing.T) {
 
 See `examples/extensions/tool-logger_test.go` for a complete example with 14 test cases covering tool calls, input handling, and session lifecycle.
 
+### Prompt Templates
+
+Create reusable prompt templates with shell-style argument substitution. Templates are loaded from `~/.kit/prompts/*.md` and `.kit/prompts/*.md`.
+
+**Example template** (`~/.kit/prompts/review.md`):
+```markdown
+---
+description: Review code for issues
+---
+Review the following code for bugs and security issues.
+Focus on $1 specifically.
+```
+
+**Usage:**
+```
+/review error handling
+```
+
+**Argument placeholders:**
+- `$1`, `$2`, etc. — Individual arguments
+- `$@` or `$ARGUMENTS` — All arguments
+- `${@:2}` — Arguments from position 2 onwards
+- `${@:1:3}` — 3 arguments starting at position 1
+
+Disable templates with `--no-prompt-templates` or load a specific template with `--prompt-template <name>`.
+
 ## Session Management
 
 Kit uses a tree-based session model that supports branching and forking conversations.
@@ -419,6 +456,22 @@ kit -s path/to/session.jsonl
 kit --no-session
 ```
 
+### Interactive Session Commands
+
+During an interactive session, use these slash commands:
+
+| Command | Description |
+|---------|-------------|
+| `/name [name]` | Set or display the session's display name |
+| `/session` | Show session info (path, ID, message count) |
+| `/resume` | Open the session picker to switch sessions |
+| `/export [path]` | Export session as JSONL (auto-generates path if omitted) |
+| `/import <path>` | Import and switch to a session from a JSONL file |
+| `/share` | Upload session to GitHub Gist and get a shareable viewer URL |
+| `/tree` | Navigate the session tree |
+| `/fork` | Branch from an earlier message |
+| `/new` | Start a fresh session |
+
 ## Go SDK
 
 Embed Kit in your Go applications:
@@ -657,7 +710,7 @@ npm/                 - NPM package wrapper for distribution
 
 ```bash
 provider/model            # Standard format
-anthropic/claude-sonnet-4-5-20250929
+anthropic/claude-sonnet-latest
 openai/gpt-4o
 ollama/llama3
 google/gemini-2.0-flash-exp
@@ -666,14 +719,35 @@ google/gemini-2.0-flash-exp
 ### Model Aliases
 
 ```bash
-claude-opus-latest        → claude-opus-4-20250514
-claude-sonnet-latest      → claude-sonnet-4-5-20250929
-claude-4-opus-latest      → claude-opus-4-20250514
-claude-4-sonnet-latest    → claude-sonnet-4-5-20250929
+# Anthropic Claude
+claude-opus-latest        → claude-opus-4-6
+claude-sonnet-latest      → claude-sonnet-4-6
+claude-haiku-latest       → claude-haiku-4-5
+claude-4-opus-latest      → claude-opus-4-6
+claude-4-sonnet-latest    → claude-sonnet-4-6
+claude-4-haiku-latest     → claude-haiku-4-5
 claude-3-7-sonnet-latest  → claude-3-7-sonnet-20250219
-claude-3-5-sonnet-latest  → claude-3-5-sonnet-20241022
+claude-3-5-sonnet-latest   → claude-3-5-sonnet-20241022
 claude-3-5-haiku-latest   → claude-3-5-haiku-20241022
 claude-3-opus-latest      → claude-3-opus-20240229
+
+# OpenAI GPT
+o1-latest                 → o1
+o3-latest                 → o3
+o4-latest                 → o4-mini
+gpt-5-latest              → gpt-5.4
+gpt-5-chat-latest         → gpt-5.4
+gpt-4-latest              → gpt-4o
+gpt-4                     → gpt-4o
+gpt-3.5-latest            → gpt-3.5-turbo
+gpt-3.5                   → gpt-3.5-turbo
+codex-latest              → codex-mini-latest
+
+# Google Gemini
+gemini-pro-latest         → gemini-2.5-pro
+gemini-flash-latest       → gemini-2.5-flash
+gemini-flash              → gemini-2.5-flash
+gemini-pro                → gemini-2.5-pro
 ```
 
 ## Contributing

cmd/root.go

@@ -16,6 +16,7 @@ import (
 	"github.com/mark3labs/kit/internal/config"
 	"github.com/mark3labs/kit/internal/extensions"
 	"github.com/mark3labs/kit/internal/models"
+	"github.com/mark3labs/kit/internal/prompts"
 	"github.com/mark3labs/kit/internal/ui"
 	kit "github.com/mark3labs/kit/pkg/kit"
 	"github.com/spf13/cobra"
@@ -65,6 +66,15 @@ var (
 
 	// TLS configuration
 	tlsSkipVerify bool
+
+	// Prompt templates
+	promptTemplatePaths []string
+	noPromptTemplates   bool
+
+	// Preference restoration flags — set in RunE after cobra parses, used
+	// in runNormalMode to decide whether to apply saved preferences.
+	modelFlagChanged    bool
+	thinkingFlagChanged bool
 )
 
 // kitUIAdapter adapts *kit.Kit to ui.AgentInterface so the CLI setup layer
@@ -113,6 +123,17 @@ var rootCmd = &cobra.Command{
 		if len(args) > 0 {
 			processPositionalArgs(args)
 		}
+		// Record whether --model / --thinking-level were explicitly set by the
+		// user so that runNormalMode can fall back to saved preferences when
+		// they weren't. Must be captured here (after cobra parses) and before
+		// runKit because rootCmd can't be referenced inside runNormalMode
+		// without creating an initialization cycle.
+		if f := cmd.PersistentFlags().Lookup("model"); f != nil {
+			modelFlagChanged = f.Changed
+		}
+		if f := cmd.PersistentFlags().Lookup("thinking-level"); f != nil {
+			thinkingFlagChanged = f.Changed
+		}
 		return runKit(context.Background())
 	},
 }
@@ -232,6 +253,9 @@ func init() {
 	if err == nil && viper.InConfig("theme") {
 		uiTheme := configToUiTheme(theme)
 		ui.SetTheme(uiTheme)
+	} else if pref := ui.LoadThemePreference(); pref != "" {
+		// No explicit theme in config — fall back to persisted preference.
+		_ = ui.ApplyThemeWithoutSave(pref)
 	}
 
 	rootCmd.PersistentFlags().
@@ -277,6 +301,10 @@ func init() {
 	flags.StringVar(&providerAPIKey, "provider-api-key", "", "API key for the provider (applies to OpenAI, Anthropic, and Google)")
 	flags.BoolVar(&tlsSkipVerify, "tls-skip-verify", false, "skip TLS certificate verification (WARNING: insecure, use only for self-signed certificates)")
 
+	// Prompt template flags
+	flags.StringArrayVar(&promptTemplatePaths, "prompt-template", nil, "load prompt template file or directory (repeatable)")
+	flags.BoolVar(&noPromptTemplates, "no-prompt-templates", false, "disable prompt template discovery")
+
 	// Model generation parameters
 	flags.IntVar(&maxTokens, "max-tokens", 4096, "maximum number of tokens in the response")
 	flags.Float32Var(&temperature, "temperature", 0.7, "controls randomness in responses (0.0-1.0)")
@@ -312,6 +340,8 @@ func init() {
 	_ = viper.BindPFlag("tls-skip-verify", rootCmd.PersistentFlags().Lookup("tls-skip-verify"))
 	_ = viper.BindPFlag("no-extensions", rootCmd.PersistentFlags().Lookup("no-extensions"))
 	_ = viper.BindPFlag("extension", rootCmd.PersistentFlags().Lookup("extension"))
+	_ = viper.BindPFlag("prompt-template", rootCmd.PersistentFlags().Lookup("prompt-template"))
+	_ = viper.BindPFlag("no-prompt-templates", rootCmd.PersistentFlags().Lookup("no-prompt-templates"))
 
 	// Defaults are already set in flag definitions, no need to duplicate in viper
 
@@ -643,6 +673,22 @@ func runNormalMode(ctx context.Context) error {
 		log.SetFlags(log.LstdFlags | log.Lshortfile)
 	}
 
+	// Restore persisted model preference when no explicit --model flag or
+	// config file model is set. Precedence: CLI flag > config file > saved
+	// preference > built-in default. This mirrors how themes are persisted.
+	if !modelFlagChanged && !viper.InConfig("model") {
+		if pref := ui.LoadModelPreference(); pref != "" {
+			viper.Set("model", pref)
+		}
+	}
+
+	// Restore persisted thinking level preference (same precedence chain).
+	if !thinkingFlagChanged && !viper.InConfig("thinking-level") {
+		if pref := ui.LoadThinkingLevelPreference(); pref != "" {
+			viper.Set("thinking-level", pref)
+		}
+	}
+
 	// Load MCP configuration.
 	mcpConfig, err := config.LoadAndValidateConfig()
 	if err != nil {
@@ -678,11 +724,16 @@ func runNormalMode(ctx context.Context) error {
 		},
 	}
 	if resumeFlag {
-		// TODO: TUI session picker.
-		sessions, _ := kit.ListSessions("")
-		if len(sessions) > 0 {
-			kitOpts.SessionPath = sessions[0].Path
+		// When --resume is combined with interactive mode, the TUI session
+		// picker will be shown at startup. For non-interactive mode, fall
+		// back to auto-selecting the most recent session.
+		if positionalPrompt != "" {
+			sessions, _ := kit.ListSessions("")
+			if len(sessions) > 0 {
+				kitOpts.SessionPath = sessions[0].Path
+			}
 		}
+		// Interactive mode: ShowSessionPicker is set below on AppModelOptions.
 	}
 
 	kitInstance, err := kit.New(ctx, kitOpts)
@@ -1024,6 +1075,27 @@ func runNormalMode(ctx context.Context) error {
 	// Convert extension commands to UI-layer type for the interactive TUI.
 	extCommands := extensionCommandsForUI(kitInstance)
 
+	// Load prompt templates from standard locations and explicit paths.
+	var promptTemplates []*prompts.PromptTemplate
+	if !noPromptTemplates {
+		homeDir, _ := os.UserHomeDir()
+		cwd, _ := os.Getwd()
+		tpls, diags, err := prompts.LoadAll(prompts.LoadOptions{
+			Cwd:             cwd,
+			HomeDir:         homeDir,
+			ExtraPaths:      promptTemplatePaths,
+			ConfigPaths:     viper.GetStringSlice("prompts"),
+			IncludeDefaults: true,
+		})
+		if err != nil {
+			log.Printf("Warning: failed to load some prompt templates: %v", err)
+		}
+		promptTemplates = tpls
+		for _, d := range diags {
+			log.Printf("Prompt template collision: /%s kept from %s, dropped from %s", d.Name, d.KeptPath, d.DroppedPath)
+		}
+	}
+
 	// Build context/skills display metadata for the startup banner.
 	var contextPaths []string
 	for _, cf := range kitInstance.GetContextFiles() {
@@ -1081,9 +1153,21 @@ func runNormalMode(ctx context.Context) error {
 		return kitInstance.SetThinkingLevel(context.Background(), level)
 	}
 
+	// Build session-switching callback. Opens a JSONL session file and
+	// replaces the active tree session on both the Kit SDK and App layer.
+	switchSessionForUI := func(path string) error {
+		ts, err := kit.OpenTreeSession(path)
+		if err != nil {
+			return fmt.Errorf("failed to open session: %w", err)
+		}
+		kitInstance.SetTreeSession(ts)
+		appInstance.SwitchTreeSession(ts)
+		return nil
+	}
+
 	// Check if running in non-interactive mode
 	if positionalPrompt != "" {
-		return runNonInteractiveModeApp(ctx, appInstance, cli, positionalPrompt, quietFlag, jsonFlag, noExitFlag, modelName, parsedProvider, kitInstance.GetLoadingMessage(), serverNames, toolNames, mcpToolCount, extensionToolCount, usageTracker, extCommands, contextPaths, skillItems, getWidgets, getHeader, getFooter, getToolRenderer, getEditorInterceptor, getUIVisibility, getStatusBarEntries, emitBeforeFork, emitBeforeSessionSwitch, getGlobalShortcuts, getExtensionCommands, setModelForUI, emitModelChangeForUI, kitInstance.IsReasoningModel(), kitInstance.GetThinkingLevel(), setThinkingLevelForUI)
+		return runNonInteractiveModeApp(ctx, appInstance, cli, positionalPrompt, quietFlag, jsonFlag, noExitFlag, modelName, parsedProvider, kitInstance.GetLoadingMessage(), serverNames, toolNames, mcpToolCount, extensionToolCount, usageTracker, extCommands, promptTemplates, contextPaths, skillItems, getWidgets, getHeader, getFooter, getToolRenderer, getEditorInterceptor, getUIVisibility, getStatusBarEntries, emitBeforeFork, emitBeforeSessionSwitch, getGlobalShortcuts, getExtensionCommands, setModelForUI, emitModelChangeForUI, kitInstance.IsReasoningModel(), kitInstance.GetThinkingLevel(), setThinkingLevelForUI, switchSessionForUI)
 	}
 
 	// Quiet mode is not allowed in interactive mode
@@ -1091,7 +1175,7 @@ func runNormalMode(ctx context.Context) error {
 		return fmt.Errorf("--quiet requires a prompt")
 	}
 
-	return runInteractiveModeBubbleTea(ctx, appInstance, modelName, parsedProvider, kitInstance.GetLoadingMessage(), serverNames, toolNames, mcpToolCount, extensionToolCount, usageTracker, extCommands, contextPaths, skillItems, getWidgets, getHeader, getFooter, getToolRenderer, getEditorInterceptor, getUIVisibility, getStatusBarEntries, emitBeforeFork, emitBeforeSessionSwitch, getGlobalShortcuts, getExtensionCommands, setModelForUI, emitModelChangeForUI, kitInstance.IsReasoningModel(), kitInstance.GetThinkingLevel(), setThinkingLevelForUI)
+	return runInteractiveModeBubbleTea(ctx, appInstance, modelName, parsedProvider, kitInstance.GetLoadingMessage(), serverNames, toolNames, mcpToolCount, extensionToolCount, usageTracker, extCommands, promptTemplates, contextPaths, skillItems, getWidgets, getHeader, getFooter, getToolRenderer, getEditorInterceptor, getUIVisibility, getStatusBarEntries, emitBeforeFork, emitBeforeSessionSwitch, getGlobalShortcuts, getExtensionCommands, setModelForUI, emitModelChangeForUI, kitInstance.IsReasoningModel(), kitInstance.GetThinkingLevel(), setThinkingLevelForUI, switchSessionForUI)
 }
 
 // runNonInteractiveModeApp executes a single prompt via the app layer and exits,
@@ -1104,7 +1188,7 @@ func runNormalMode(ctx context.Context) error {
 //
 // When --no-exit is set, after the prompt completes the interactive BubbleTea
 // TUI is started so the user can continue the conversation.
-func runNonInteractiveModeApp(ctx context.Context, appInstance *app.App, cli *ui.CLI, prompt string, quiet, jsonOutput, noExit bool, modelName, providerName, loadingMessage string, serverNames, toolNames []string, mcpToolCount, extensionToolCount int, usageTracker *ui.UsageTracker, extCommands []ui.ExtensionCommand, contextPaths []string, skillItems []ui.SkillItem, getWidgets func(string) []ui.WidgetData, getHeader, getFooter func() *ui.WidgetData, getToolRenderer func(string) *ui.ToolRendererData, getEditorInterceptor func() *ui.EditorInterceptor, getUIVisibility func() *ui.UIVisibility, getStatusBarEntries func() []ui.StatusBarEntryData, emitBeforeFork func(string, bool, string) (bool, string), emitBeforeSessionSwitch func(string) (bool, string), getGlobalShortcuts func() map[string]func(), getExtensionCommands func() []ui.ExtensionCommand, setModel func(string) error, emitModelChange func(string, string, string), isReasoningModel bool, thinkingLevel string, setThinkingLevel func(string) error) error {
+func runNonInteractiveModeApp(ctx context.Context, appInstance *app.App, cli *ui.CLI, prompt string, quiet, jsonOutput, noExit bool, modelName, providerName, loadingMessage string, serverNames, toolNames []string, mcpToolCount, extensionToolCount int, usageTracker *ui.UsageTracker, extCommands []ui.ExtensionCommand, promptTemplates []*prompts.PromptTemplate, contextPaths []string, skillItems []ui.SkillItem, getWidgets func(string) []ui.WidgetData, getHeader, getFooter func() *ui.WidgetData, getToolRenderer func(string) *ui.ToolRendererData, getEditorInterceptor func() *ui.EditorInterceptor, getUIVisibility func() *ui.UIVisibility, getStatusBarEntries func() []ui.StatusBarEntryData, emitBeforeFork func(string, bool, string) (bool, string), emitBeforeSessionSwitch func(string) (bool, string), getGlobalShortcuts func() map[string]func(), getExtensionCommands func() []ui.ExtensionCommand, setModel func(string) error, emitModelChange func(string, string, string), isReasoningModel bool, thinkingLevel string, setThinkingLevel func(string) error, switchSession func(string) error) error {
 	// Expand @file references in the prompt before sending to the agent.
 	if cwd, err := os.Getwd(); err == nil {
 		prompt = ui.ProcessFileAttachments(prompt, cwd)
@@ -1147,7 +1231,7 @@ func runNonInteractiveModeApp(ctx context.Context, appInstance *app.App, cli *ui
 
 	// If --no-exit was requested, hand off to the interactive TUI.
 	if noExit {
-		return runInteractiveModeBubbleTea(ctx, appInstance, modelName, providerName, loadingMessage, serverNames, toolNames, mcpToolCount, extensionToolCount, usageTracker, extCommands, contextPaths, skillItems, getWidgets, getHeader, getFooter, getToolRenderer, getEditorInterceptor, getUIVisibility, getStatusBarEntries, emitBeforeFork, emitBeforeSessionSwitch, getGlobalShortcuts, getExtensionCommands, setModel, emitModelChange, isReasoningModel, thinkingLevel, setThinkingLevel)
+		return runInteractiveModeBubbleTea(ctx, appInstance, modelName, providerName, loadingMessage, serverNames, toolNames, mcpToolCount, extensionToolCount, usageTracker, extCommands, promptTemplates, contextPaths, skillItems, getWidgets, getHeader, getFooter, getToolRenderer, getEditorInterceptor, getUIVisibility, getStatusBarEntries, emitBeforeFork, emitBeforeSessionSwitch, getGlobalShortcuts, getExtensionCommands, setModel, emitModelChange, isReasoningModel, thinkingLevel, setThinkingLevel, switchSession)
 	}
 
 	return nil
@@ -1245,7 +1329,7 @@ func writeJSONError(err error) {
 //  4. Calls program.Run() which blocks until the user quits (Ctrl+C or /quit).
 //
 // SetupCLI is not used for interactive mode; the TUI (AppModel) handles its own rendering.
-func runInteractiveModeBubbleTea(_ context.Context, appInstance *app.App, modelName, providerName, loadingMessage string, serverNames, toolNames []string, mcpToolCount, extensionToolCount int, usageTracker *ui.UsageTracker, extCommands []ui.ExtensionCommand, contextPaths []string, skillItems []ui.SkillItem, getWidgets func(string) []ui.WidgetData, getHeader, getFooter func() *ui.WidgetData, getToolRenderer func(string) *ui.ToolRendererData, getEditorInterceptor func() *ui.EditorInterceptor, getUIVisibility func() *ui.UIVisibility, getStatusBarEntries func() []ui.StatusBarEntryData, emitBeforeFork func(string, bool, string) (bool, string), emitBeforeSessionSwitch func(string) (bool, string), getGlobalShortcuts func() map[string]func(), getExtensionCommands func() []ui.ExtensionCommand, setModel func(string) error, emitModelChange func(string, string, string), isReasoningModel bool, thinkingLevel string, setThinkingLevel func(string) error) error {
+func runInteractiveModeBubbleTea(_ context.Context, appInstance *app.App, modelName, providerName, loadingMessage string, serverNames, toolNames []string, mcpToolCount, extensionToolCount int, usageTracker *ui.UsageTracker, extCommands []ui.ExtensionCommand, promptTemplates []*prompts.PromptTemplate, contextPaths []string, skillItems []ui.SkillItem, getWidgets func(string) []ui.WidgetData, getHeader, getFooter func() *ui.WidgetData, getToolRenderer func(string) *ui.ToolRendererData, getEditorInterceptor func() *ui.EditorInterceptor, getUIVisibility func() *ui.UIVisibility, getStatusBarEntries func() []ui.StatusBarEntryData, emitBeforeFork func(string, bool, string) (bool, string), emitBeforeSessionSwitch func(string) (bool, string), getGlobalShortcuts func() map[string]func(), getExtensionCommands func() []ui.ExtensionCommand, setModel func(string) error, emitModelChange func(string, string, string), isReasoningModel bool, thinkingLevel string, setThinkingLevel func(string) error, switchSession func(string) error) error {
 	// Determine terminal size; fall back gracefully.
 	termWidth, termHeight, err := term.GetSize(int(os.Stdout.Fd()))
 	if err != nil || termWidth == 0 {
@@ -1254,6 +1338,7 @@ func runInteractiveModeBubbleTea(_ context.Context, appInstance *app.App, modelN
 	}
 
 	cwd, _ := os.Getwd()
+
 	appModel := ui.NewAppModel(appInstance, ui.AppModelOptions{
 		CompactMode:             viper.GetBool("compact"),
 		ModelName:               modelName,
@@ -1268,6 +1353,7 @@ func runInteractiveModeBubbleTea(_ context.Context, appInstance *app.App, modelN
 		ExtensionToolCount:      extensionToolCount,
 		UsageTracker:            usageTracker,
 		ExtensionCommands:       extCommands,
+		PromptTemplates:         promptTemplates,
 		ContextPaths:            contextPaths,
 		SkillItems:              skillItems,
 		GetWidgets:              getWidgets,
@@ -1286,6 +1372,8 @@ func runInteractiveModeBubbleTea(_ context.Context, appInstance *app.App, modelN
 		ThinkingLevel:           thinkingLevel,
 		IsReasoningModel:        isReasoningModel,
 		SetThinkingLevel:        setThinkingLevel,
+		SwitchSession:           switchSession,
+		ShowSessionPicker:       resumeFlag,
 	})
 
 	// Print startup info to stdout before Bubble Tea takes over the screen.

cmd/skill.go

@@ -8,19 +8,21 @@ import (
 	"github.com/spf13/cobra"
 )
 
-// skillCmd installs the kit-extensions skill via the skills.sh CLI (npx skills).
-// This teaches AI agents how to create Kit extensions with full knowledge of
-// the extension API, lifecycle events, widgets, tools, commands, and Yaegi constraints.
+// skillCmd installs Kit skills via the skills.sh CLI (npx skills).
 var skillCmd = &cobra.Command{
 	Use:   "skill",
-	Short: "Install the Kit extensions skill via skills.sh",
-	Long: `Install the kit-extensions skill that teaches AI agents how to create
-Kit extensions. Uses the skills.sh CLI (npx skills) to install the skill
-from the Kit repository.
+	Short: "Install Kit skills via skills.sh",
+	Long: `Install Kit skills that teach AI agents how to build with Kit.
+Uses the skills.sh CLI (npx skills) to install all skills from the Kit repository.
 
-The skill provides comprehensive documentation of Kit's extension API including
-lifecycle events, custom tools, slash commands, widgets, editor interceptors,
-tool renderers, and critical Yaegi interpreter constraints.
+Two skills are provided:
+
+  1. Extensions — creating Kit extensions with full knowledge of the extension
+     API, lifecycle events, widgets, tools, commands, editor interceptors,
+     tool renderers, and Yaegi interpreter constraints.
+
+  2. SDK — building AI-powered applications with the Kit Go SDK, including
+     providers, agents, tools, and MCP integration.
 
 Example:
   kit skill`,
@@ -41,8 +43,6 @@ func runSkill(_ *cobra.Command, _ []string) error {
 		"skills",
 		"add",
 		"mark3labs/kit",
-		"--skill",
-		"kit-extensions",
 	}
 
 	cmd := exec.Command(npx, args...)

examples/sdk/README.md

@@ -0,0 +1,45 @@
+# SDK Examples
+
+These examples demonstrate how to use the Kit SDK (`pkg/kit`) to build agents programmatically in Go.
+
+## Examples
+
+### [basic](basic/)
+
+Shows core SDK usage: creating a Kit instance, sending prompts, overriding the model, subscribing to events (tool calls, streaming), and session management.
+
+```bash
+go run ./examples/sdk/basic
+```
+
+### [scripting](scripting/)
+
+A minimal script-friendly wrapper that takes a prompt from the command line and prints the response — useful for piping and automation.
+
+```bash
+go run ./examples/sdk/scripting "Explain what this repo does"
+```
+
+### [crypto-monitor](crypto-monitor/)
+
+A background agent that checks Bitcoin and Ethereum prices every 30 minutes and sends desktop notifications via `notify-send` (dbus). Demonstrates using the SDK for a long-running autonomous task with a single tool.
+
+```bash
+go run ./examples/sdk/crypto-monitor
+
+# Override the check interval:
+CRYPTO_INTERVAL=5m go run ./examples/sdk/crypto-monitor
+```
+
+## Getting Started
+
+```go
+import kit "github.com/mark3labs/kit/pkg/kit"
+
+host, err := kit.New(ctx, nil)        // uses ~/.kit.yml defaults
+defer host.Close()
+
+response, err := host.Prompt(ctx, "Hello!")
+```
+
+See the [SDK README](../../pkg/kit/README.md) for the full API reference.

examples/sdk/crypto-monitor/main.go

@@ -0,0 +1,85 @@
+package main
+
+import (
+	"context"
+	"fmt"
+	"log"
+	"os"
+	"os/signal"
+	"time"
+
+	kit "github.com/mark3labs/kit/pkg/kit"
+)
+
+const systemPrompt = `You are a cryptocurrency price monitor. Your job is to:
+
+1. Fetch the current prices of Bitcoin and Ethereum using bash with curl
+2. Send a desktop notification with the results using notify-send
+
+To fetch prices, use this CoinGecko API endpoint (no API key needed):
+  curl -s 'https://api.coingecko.com/api/v3/simple/price?ids=bitcoin,ethereum&vs_currencies=usd&include_24hr_change=true'
+
+To send a desktop notification:
+  notify-send -i dialog-information "Crypto Prices" "BTC: $XX,XXX (+X.X%)\nETH: $X,XXX (+X.X%)"
+
+Include the 24h percentage change in the notification. Use a green arrow (▲) for
+positive changes and a red arrow (▼) for negative. Format prices with commas.
+
+If the API call fails, send a notification about the failure instead.
+
+Always complete both steps: fetch then notify. Be concise — no commentary needed.`
+
+func main() {
+	interval := 30 * time.Minute
+	if os.Getenv("CRYPTO_INTERVAL") != "" {
+		d, err := time.ParseDuration(os.Getenv("CRYPTO_INTERVAL"))
+		if err == nil {
+			interval = d
+		}
+	}
+
+	ctx, cancel := signal.NotifyContext(context.Background(), os.Interrupt)
+	defer cancel()
+
+	host, err := kit.New(ctx, &kit.Options{
+		SystemPrompt: systemPrompt,
+		Tools:        []kit.Tool{kit.NewBashTool()},
+		NoSession:    true,
+		Quiet:        true,
+	})
+	if err != nil {
+		log.Fatalf("Failed to create kit instance: %v", err)
+	}
+	defer func() { _ = host.Close() }()
+
+	fmt.Printf("Crypto price monitor started (every %s)\n", interval)
+	fmt.Println("Press Ctrl+C to stop")
+
+	// Run immediately on startup, then on each tick.
+	check(ctx, host)
+
+	ticker := time.NewTicker(interval)
+	defer ticker.Stop()
+
+	for {
+		select {
+		case <-ticker.C:
+			check(ctx, host)
+		case <-ctx.Done():
+			fmt.Println("\nStopping price monitor")
+			return
+		}
+	}
+}
+
+func check(ctx context.Context, host *kit.Kit) {
+	fmt.Printf("[%s] Checking prices...\n", time.Now().Format("15:04:05"))
+
+	// Clear session so each check is independent.
+	host.ClearSession()
+
+	_, err := host.Prompt(ctx, "Fetch current Bitcoin and Ethereum prices and send a desktop notification.")
+	if err != nil {
+		fmt.Fprintf(os.Stderr, "Error: %v\n", err)
+	}
+}

go.mod

@@ -1,26 +1,27 @@
 module github.com/mark3labs/kit
 
-go 1.26.0
+go 1.26.1
 
 require (
 	charm.land/bubbles/v2 v2.0.0
 	charm.land/bubbletea/v2 v2.0.2
-	charm.land/fantasy v0.11.1
-	charm.land/lipgloss/v2 v2.0.1
+	charm.land/fantasy v0.16.0
+	charm.land/huh/v2 v2.0.3
+	charm.land/lipgloss/v2 v2.0.2
 	github.com/alecthomas/chroma/v2 v2.23.1
 	github.com/aymanbagabas/go-udiff v0.4.1
-	github.com/charmbracelet/fang v0.4.4
-	github.com/charmbracelet/log v0.4.2
-	github.com/mark3labs/mcp-go v0.44.1
+	github.com/charmbracelet/fang v1.0.0
+	github.com/charmbracelet/log v1.0.0
+	github.com/coder/acp-go-sdk v0.6.3
+	github.com/mark3labs/mcp-go v0.45.0
 	github.com/spf13/cobra v1.10.2
 	github.com/spf13/viper v1.21.0
 	github.com/traefik/yaegi v0.16.1
-	golang.org/x/term v0.40.0
+	golang.org/x/term v0.41.0
 	gopkg.in/yaml.v3 v3.0.1
 )
 
 require (
-	charm.land/huh/v2 v2.0.3 // indirect
 	cloud.google.com/go v0.123.0 // indirect
 	cloud.google.com/go/auth v0.18.2 // indirect
 	cloud.google.com/go/auth/oauth2adapt v0.2.8 // indirect
@@ -28,42 +29,42 @@ require (
 	github.com/Azure/azure-sdk-for-go/sdk/azcore v1.21.0 // indirect
 	github.com/Azure/azure-sdk-for-go/sdk/internal v1.11.2 // indirect
 	github.com/atotto/clipboard v0.1.4 // indirect
-	github.com/aws/aws-sdk-go-v2 v1.41.3 // indirect
-	github.com/aws/aws-sdk-go-v2/aws/protocol/eventstream v1.7.6 // indirect
-	github.com/aws/aws-sdk-go-v2/config v1.32.11 // indirect
-	github.com/aws/aws-sdk-go-v2/credentials v1.19.11 // indirect
-	github.com/aws/aws-sdk-go-v2/feature/ec2/imds v1.18.19 // indirect
-	github.com/aws/aws-sdk-go-v2/internal/configsources v1.4.19 // indirect
-	github.com/aws/aws-sdk-go-v2/internal/endpoints/v2 v2.7.19 // indirect
-	github.com/aws/aws-sdk-go-v2/internal/ini v1.8.5 // indirect
-	github.com/aws/aws-sdk-go-v2/service/internal/accept-encoding v1.13.6 // indirect
-	github.com/aws/aws-sdk-go-v2/service/internal/presigned-url v1.13.19 // indirect
-	github.com/aws/aws-sdk-go-v2/service/signin v1.0.7 // indirect
-	github.com/aws/aws-sdk-go-v2/service/sso v1.30.12 // indirect
-	github.com/aws/aws-sdk-go-v2/service/ssooidc v1.35.16 // indirect
-	github.com/aws/aws-sdk-go-v2/service/sts v1.41.8 // indirect
+	github.com/aws/aws-sdk-go-v2 v1.41.4 // indirect
+	github.com/aws/aws-sdk-go-v2/aws/protocol/eventstream v1.7.7 // indirect
+	github.com/aws/aws-sdk-go-v2/config v1.32.12 // indirect
+	github.com/aws/aws-sdk-go-v2/credentials v1.19.12 // indirect
+	github.com/aws/aws-sdk-go-v2/feature/ec2/imds v1.18.20 // indirect
+	github.com/aws/aws-sdk-go-v2/internal/configsources v1.4.20 // indirect
+	github.com/aws/aws-sdk-go-v2/internal/endpoints/v2 v2.7.20 // indirect
+	github.com/aws/aws-sdk-go-v2/internal/ini v1.8.6 // indirect
+	github.com/aws/aws-sdk-go-v2/service/internal/accept-encoding v1.13.7 // indirect
+	github.com/aws/aws-sdk-go-v2/service/internal/presigned-url v1.13.20 // indirect
+	github.com/aws/aws-sdk-go-v2/service/signin v1.0.8 // indirect
+	github.com/aws/aws-sdk-go-v2/service/sso v1.30.13 // indirect
+	github.com/aws/aws-sdk-go-v2/service/ssooidc v1.35.17 // indirect
+	github.com/aws/aws-sdk-go-v2/service/sts v1.41.9 // indirect
 	github.com/aws/smithy-go v1.24.2 // indirect
 	github.com/aymerick/douceur v0.2.0 // indirect
 	github.com/bahlo/generic-list-go v0.2.0 // indirect
-	github.com/buger/jsonparser v1.1.1 // indirect
-	github.com/catppuccin/go v0.2.0 // indirect
+	github.com/buger/jsonparser v1.1.2 // indirect
+	github.com/catppuccin/go v0.3.0 // indirect
 	github.com/cespare/xxhash/v2 v2.3.0 // indirect
 	github.com/charmbracelet/anthropic-sdk-go v0.0.0-20260223140439-63879b0b8dab // indirect
-	github.com/charmbracelet/colorprofile v0.4.2 // indirect
+	github.com/charmbracelet/colorprofile v0.4.3 // indirect
 	github.com/charmbracelet/harmonica v0.2.0 // indirect
 	github.com/charmbracelet/lipgloss v1.1.1-0.20250404203927-76690c660834 // indirect
-	github.com/charmbracelet/ultraviolet v0.0.0-20260303162955-0b88c25f3fff // indirect
+	github.com/charmbracelet/openai-go v0.0.0-20260319145158-d0740cc34266 // indirect
+	github.com/charmbracelet/ultraviolet v0.0.0-20260316091819-b93f6a3b8502 // indirect
 	github.com/charmbracelet/x/cellbuf v0.0.15 // indirect
-	github.com/charmbracelet/x/exp/charmtone v0.0.0-20260305213658-fe36e8c10185 // indirect
+	github.com/charmbracelet/x/exp/charmtone v0.0.0-20260322003602-9b007323c5cd // indirect
 	github.com/charmbracelet/x/exp/ordered v0.1.0 // indirect
-	github.com/charmbracelet/x/exp/slice v0.0.0-20260305213658-fe36e8c10185 // indirect
-	github.com/charmbracelet/x/exp/strings v0.0.0-20240722160745-212f7b056ed0 // indirect
+	github.com/charmbracelet/x/exp/slice v0.0.0-20260322003602-9b007323c5cd // indirect
+	github.com/charmbracelet/x/exp/strings v0.1.0 // indirect
 	github.com/charmbracelet/x/json v0.2.0 // indirect
 	github.com/charmbracelet/x/termios v0.1.1 // indirect
 	github.com/charmbracelet/x/windows v0.2.2 // indirect
 	github.com/clipperhouse/displaywidth v0.11.0 // indirect
 	github.com/clipperhouse/uax29/v2 v2.7.0 // indirect
-	github.com/coder/acp-go-sdk v0.6.3 // indirect
 	github.com/dlclark/regexp2 v1.11.5 // indirect
 	github.com/dustin/go-humanize v1.0.1 // indirect
 	github.com/felixge/httpsnoop v1.0.4 // indirect
@@ -79,15 +80,15 @@ require (
 	github.com/google/s2a-go v0.1.9 // indirect
 	github.com/google/uuid v1.6.0 // indirect
 	github.com/googleapis/enterprise-certificate-proxy v0.3.14 // indirect
-	github.com/googleapis/gax-go/v2 v2.17.0 // indirect
+	github.com/googleapis/gax-go/v2 v2.19.0 // indirect
 	github.com/gorilla/css v1.0.1 // indirect
 	github.com/gorilla/websocket v1.5.3 // indirect
 	github.com/invopop/jsonschema v0.13.0 // indirect
 	github.com/kaptinlin/go-i18n v0.2.12 // indirect
 	github.com/kaptinlin/jsonpointer v0.4.17 // indirect
-	github.com/kaptinlin/jsonschema v0.7.5 // indirect
+	github.com/kaptinlin/jsonschema v0.7.6 // indirect
 	github.com/kaptinlin/messageformat-go v0.4.18 // indirect
-	github.com/mailru/easyjson v0.9.1 // indirect
+	github.com/mailru/easyjson v0.9.2 // indirect
 	github.com/microcosm-cc/bluemonday v1.0.27 // indirect
 	github.com/mitchellh/hashstructure/v2 v2.0.2 // indirect
 	github.com/muesli/mango v0.2.0 // indirect
@@ -95,7 +96,6 @@ require (
 	github.com/muesli/mango-pflag v0.2.0 // indirect
 	github.com/muesli/reflow v0.3.0 // indirect
 	github.com/muesli/roff v0.1.0 // indirect
-	github.com/openai/openai-go/v2 v2.7.1 // indirect
 	github.com/pelletier/go-toml/v2 v2.2.4 // indirect
 	github.com/sagikazarmark/locafero v0.12.0 // indirect
 	github.com/spf13/afero v1.15.0 // indirect
@@ -108,42 +108,42 @@ require (
 	github.com/wk8/go-ordered-map/v2 v2.1.8 // indirect
 	github.com/xo/terminfo v0.0.0-20220910002029-abceb7e1c41e // indirect
 	github.com/yosida95/uritemplate/v3 v3.0.2 // indirect
-	github.com/yuin/goldmark v1.7.16 // indirect
+	github.com/yuin/goldmark v1.7.17 // indirect
 	github.com/yuin/goldmark-emoji v1.0.6 // indirect
 	go.opentelemetry.io/auto/sdk v1.2.1 // indirect
-	go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc v0.66.0 // indirect
-	go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.66.0 // indirect
-	go.opentelemetry.io/otel v1.41.0 // indirect
-	go.opentelemetry.io/otel/metric v1.41.0 // indirect
-	go.opentelemetry.io/otel/trace v1.41.0 // indirect
+	go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc v0.67.0 // indirect
+	go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.67.0 // indirect
+	go.opentelemetry.io/otel v1.42.0 // indirect
+	go.opentelemetry.io/otel/metric v1.42.0 // indirect
+	go.opentelemetry.io/otel/trace v1.42.0 // indirect
 	go.yaml.in/yaml/v3 v3.0.4 // indirect
-	golang.org/x/crypto v0.48.0 // indirect
-	golang.org/x/exp v0.0.0-20260218203240-3dfff04db8fa // indirect
-	golang.org/x/net v0.51.0 // indirect
-	golang.org/x/oauth2 v0.35.0 // indirect
-	golang.org/x/time v0.14.0 // indirect
-	google.golang.org/api v0.269.0 // indirect
-	google.golang.org/genai v1.49.0 // indirect
-	google.golang.org/genproto/googleapis/rpc v0.0.0-20260226221140-a57be14db171 // indirect
-	google.golang.org/grpc v1.79.2 // indirect
+	golang.org/x/crypto v0.49.0 // indirect
+	golang.org/x/exp v0.0.0-20260312153236-7ab1446f8b90 // indirect
+	golang.org/x/net v0.52.0 // indirect
+	golang.org/x/oauth2 v0.36.0 // indirect
+	golang.org/x/time v0.15.0 // indirect
+	google.golang.org/api v0.272.0 // indirect
+	google.golang.org/genai v1.51.0 // indirect
+	google.golang.org/genproto/googleapis/rpc v0.0.0-20260319201613-d00831a3d3e7 // indirect
+	google.golang.org/grpc v1.79.3 // indirect
 	google.golang.org/protobuf v1.36.11 // indirect
 	gopkg.in/yaml.v2 v2.4.0 // indirect
 )
 
 require (
 	github.com/aymanbagabas/go-osc52/v2 v2.0.1 // indirect
-	github.com/charmbracelet/glamour v0.10.0
-	github.com/charmbracelet/x/ansi v0.11.6 // indirect
+	github.com/charmbracelet/glamour v1.0.0
+	github.com/charmbracelet/x/ansi v0.11.6
 	github.com/charmbracelet/x/term v0.2.2 // indirect
 	github.com/inconshreveable/mousetrap v1.1.0 // indirect
 	github.com/lucasb-eyer/go-colorful v1.3.0 // indirect
 	github.com/mattn/go-isatty v0.0.20 // indirect
-	github.com/mattn/go-runewidth v0.0.20 // indirect
+	github.com/mattn/go-runewidth v0.0.21 // indirect
 	github.com/muesli/cancelreader v0.2.2 // indirect
 	github.com/muesli/termenv v0.16.0 // indirect
 	github.com/rivo/uniseg v0.4.7 // indirect
 	github.com/spf13/pflag v1.0.10 // indirect
-	golang.org/x/sync v0.19.0 // indirect
+	golang.org/x/sync v0.20.0 // indirect
 	golang.org/x/sys v0.42.0 // indirect
-	golang.org/x/text v0.34.0 // indirect
+	golang.org/x/text v0.35.0
 )

go.sum

@@ -1,17 +1,13 @@
 charm.land/bubbles/v2 v2.0.0 h1:tE3eK/pHjmtrDiRdoC9uGNLgpopOd8fjhEe31B/ai5s=
 charm.land/bubbles/v2 v2.0.0/go.mod h1:rCHoleP2XhU8um45NTuOWBPNVHxnkXKTiZqcclL/qOI=
-charm.land/bubbletea/v2 v2.0.1 h1:B8e9zzK7x9JJ+XvHGF4xnYu9Xa0E0y0MyggY6dbaCfQ=
-charm.land/bubbletea/v2 v2.0.1/go.mod h1:3LRff2U4WIYXy7MTxfbAQ+AdfM3D8Xuvz2wbsOD9OHQ=
 charm.land/bubbletea/v2 v2.0.2 h1:4CRtRnuZOdFDTWSff9r8QFt/9+z6Emubz3aDMnf/dx0=
 charm.land/bubbletea/v2 v2.0.2/go.mod h1:3LRff2U4WIYXy7MTxfbAQ+AdfM3D8Xuvz2wbsOD9OHQ=
-charm.land/fantasy v0.11.1 h1:G1dRqkzEQ0RJN1Ls5mte8HOi0wFKxYd5bfnRAmeYvDk=
-charm.land/fantasy v0.11.1/go.mod h1:C8wNxWlw+b2z54zsTor9r1tG2GE2C4QotvAlgXh9KF8=
+charm.land/fantasy v0.16.0 h1:vE/6sR9nPcSD8qXJXX6wR8NXjtWlBVAzwQmTh5pHVrs=
+charm.land/fantasy v0.16.0/go.mod h1:VZjpXVh7IgeiIzGQybEnKzd68ofDsRj94+kzH1ZCAfQ=
 charm.land/huh/v2 v2.0.3 h1:2cJsMqEPwSywGHvdlKsJyQKPtSJLVnFKyFbsYZTlLkU=
 charm.land/huh/v2 v2.0.3/go.mod h1:93eEveeeqn47MwiC3tf+2atZ2l7Is88rAtmZNZ8x9Wc=
-charm.land/lipgloss/v2 v2.0.0 h1:sd8N/B3x892oiOjFfBQdXBQp3cAkvjGaU5TvVZC3ivo=
-charm.land/lipgloss/v2 v2.0.0/go.mod h1:w6SnmsBFBmEFBodiEDurGS/sdUY/u1+v72DqUzc6J14=
-charm.land/lipgloss/v2 v2.0.1 h1:6Xzrn49+Py1Um5q/wZG1gWgER2+7dUyZ9XMEufqPSys=
-charm.land/lipgloss/v2 v2.0.1/go.mod h1:KjPle2Qd3YmvP1KL5OMHiHysGcNwq6u83MUjYkFvEkM=
+charm.land/lipgloss/v2 v2.0.2 h1:xFolbF8JdpNkM2cEPTfXEcW1p6NRzOWTSamRfYEw8cs=
+charm.land/lipgloss/v2 v2.0.2/go.mod h1:KjPle2Qd3YmvP1KL5OMHiHysGcNwq6u83MUjYkFvEkM=
 cloud.google.com/go v0.123.0 h1:2NAUJwPR47q+E35uaJeYoNhuNEM9kM8SjgRgdeOJUSE=
 cloud.google.com/go v0.123.0/go.mod h1:xBoMV08QcqUGuPW65Qfm1o9Y4zKZBpGS+7bImXLTAZU=
 cloud.google.com/go/auth v0.18.2 h1:+Nbt5Ev0xEqxlNjd6c+yYUeosQ5TtEUaNcN/3FozlaM=
@@ -38,82 +34,86 @@ github.com/alecthomas/repr v0.5.2 h1:SU73FTI9D1P5UNtvseffFSGmdNci/O6RsqzeXJtP0Qs
 github.com/alecthomas/repr v0.5.2/go.mod h1:Fr0507jx4eOXV7AlPV6AVZLYrLIuIeSOWtW57eE/O/4=
 github.com/atotto/clipboard v0.1.4 h1:EH0zSVneZPSuFR11BlR9YppQTVDbh5+16AmcJi4g1z4=
 github.com/atotto/clipboard v0.1.4/go.mod h1:ZY9tmq7sm5xIbd9bOK4onWV4S6X0u6GY7Vn0Yu86PYI=
-github.com/aws/aws-sdk-go-v2 v1.41.3 h1:4kQ/fa22KjDt13QCy1+bYADvdgcxpfH18f0zP542kZA=
-github.com/aws/aws-sdk-go-v2 v1.41.3/go.mod h1:mwsPRE8ceUUpiTgF7QmQIJ7lgsKUPQOUl3o72QBrE1o=
-github.com/aws/aws-sdk-go-v2/aws/protocol/eventstream v1.7.6 h1:N4lRUXZpZ1KVEUn6hxtco/1d2lgYhNn1fHkkl8WhlyQ=
-github.com/aws/aws-sdk-go-v2/aws/protocol/eventstream v1.7.6/go.mod h1:lyw7GFp3qENLh7kwzf7iMzAxDn+NzjXEAGjKS2UOKqI=
-github.com/aws/aws-sdk-go-v2/config v1.32.11 h1:ftxI5sgz8jZkckuUHXfC/wMUc8u3fG1vQS0plr2F2Zs=
-github.com/aws/aws-sdk-go-v2/config v1.32.11/go.mod h1:twF11+6ps9aNRKEDimksp923o44w/Thk9+8YIlzWMmo=
-github.com/aws/aws-sdk-go-v2/credentials v1.19.11 h1:NdV8cwCcAXrCWyxArt58BrvZJ9pZ9Fhf9w6Uh5W3Uyc=
-github.com/aws/aws-sdk-go-v2/credentials v1.19.11/go.mod h1:30yY2zqkMPdrvxBqzI9xQCM+WrlrZKSOpSJEsylVU+8=
-github.com/aws/aws-sdk-go-v2/feature/ec2/imds v1.18.19 h1:INUvJxmhdEbVulJYHI061k4TVuS3jzzthNvjqvVvTKM=
-github.com/aws/aws-sdk-go-v2/feature/ec2/imds v1.18.19/go.mod h1:FpZN2QISLdEBWkayloda+sZjVJL+e9Gl0k1SyTgcswU=
-github.com/aws/aws-sdk-go-v2/internal/configsources v1.4.19 h1:/sECfyq2JTifMI2JPyZ4bdRN77zJmr6SrS1eL3augIA=
-github.com/aws/aws-sdk-go-v2/internal/configsources v1.4.19/go.mod h1:dMf8A5oAqr9/oxOfLkC/c2LU/uMcALP0Rgn2BD5LWn0=
-github.com/aws/aws-sdk-go-v2/internal/endpoints/v2 v2.7.19 h1:AWeJMk33GTBf6J20XJe6qZoRSJo0WfUhsMdUKhoODXE=
-github.com/aws/aws-sdk-go-v2/internal/endpoints/v2 v2.7.19/go.mod h1:+GWrYoaAsV7/4pNHpwh1kiNLXkKaSoppxQq9lbH8Ejw=
-github.com/aws/aws-sdk-go-v2/internal/ini v1.8.5 h1:clHU5fm//kWS1C2HgtgWxfQbFbx4b6rx+5jzhgX9HrI=
-github.com/aws/aws-sdk-go-v2/internal/ini v1.8.5/go.mod h1:O3h0IK87yXci+kg6flUKzJnWeziQUKciKrLjcatSNcY=
-github.com/aws/aws-sdk-go-v2/service/internal/accept-encoding v1.13.6 h1:XAq62tBTJP/85lFD5oqOOe7YYgWxY9LvWq8plyDvDVg=
-github.com/aws/aws-sdk-go-v2/service/internal/accept-encoding v1.13.6/go.mod h1:x0nZssQ3qZSnIcePWLvcoFisRXJzcTVvYpAAdYX8+GI=
-github.com/aws/aws-sdk-go-v2/service/internal/presigned-url v1.13.19 h1:X1Tow7suZk9UCJHE1Iw9GMZJJl0dAnKXXP1NaSDHwmw=
-github.com/aws/aws-sdk-go-v2/service/internal/presigned-url v1.13.19/go.mod h1:/rARO8psX+4sfjUQXp5LLifjUt8DuATZ31WptNJTyQA=
-github.com/aws/aws-sdk-go-v2/service/signin v1.0.7 h1:Y2cAXlClHsXkkOvWZFXATr34b0hxxloeQu/pAZz2row=
-github.com/aws/aws-sdk-go-v2/service/signin v1.0.7/go.mod h1:idzZ7gmDeqeNrSPkdbtMp9qWMgcBwykA7P7Rzh5DXVU=
-github.com/aws/aws-sdk-go-v2/service/sso v1.30.12 h1:iSsvB9EtQ09YrsmIc44Heqlx5ByGErqhPK1ZQLppias=
-github.com/aws/aws-sdk-go-v2/service/sso v1.30.12/go.mod h1:fEWYKTRGoZNl8tZ77i61/ccwOMJdGxwOhWCkp6TXAr0=
-github.com/aws/aws-sdk-go-v2/service/ssooidc v1.35.16 h1:EnUdUqRP1CNzt2DkV67tJx6XDN4xlfBFm+bzeNOQVb0=
-github.com/aws/aws-sdk-go-v2/service/ssooidc v1.35.16/go.mod h1:Jic/xv0Rq/pFNCh3WwpH4BEqdbSAl+IyHro8LbibHD8=
-github.com/aws/aws-sdk-go-v2/service/sts v1.41.8 h1:XQTQTF75vnug2TXS8m7CVJfC2nniYPZnO1D4Np761Oo=
-github.com/aws/aws-sdk-go-v2/service/sts v1.41.8/go.mod h1:Xgx+PR1NUOjNmQY+tRMnouRp83JRM8pRMw/vCaVhPkI=
+github.com/aws/aws-sdk-go-v2 v1.41.4 h1:10f50G7WyU02T56ox1wWXq+zTX9I1zxG46HYuG1hH/k=
+github.com/aws/aws-sdk-go-v2 v1.41.4/go.mod h1:mwsPRE8ceUUpiTgF7QmQIJ7lgsKUPQOUl3o72QBrE1o=
+github.com/aws/aws-sdk-go-v2/aws/protocol/eventstream v1.7.7 h1:3kGOqnh1pPeddVa/E37XNTaWJ8W6vrbYV9lJEkCnhuY=
+github.com/aws/aws-sdk-go-v2/aws/protocol/eventstream v1.7.7/go.mod h1:lyw7GFp3qENLh7kwzf7iMzAxDn+NzjXEAGjKS2UOKqI=
+github.com/aws/aws-sdk-go-v2/config v1.32.12 h1:O3csC7HUGn2895eNrLytOJQdoL2xyJy0iYXhoZ1OmP0=
+github.com/aws/aws-sdk-go-v2/config v1.32.12/go.mod h1:96zTvoOFR4FURjI+/5wY1vc1ABceROO4lWgWJuxgy0g=
+github.com/aws/aws-sdk-go-v2/credentials v1.19.12 h1:oqtA6v+y5fZg//tcTWahyN9PEn5eDU/Wpvc2+kJ4aY8=
+github.com/aws/aws-sdk-go-v2/credentials v1.19.12/go.mod h1:U3R1RtSHx6NB0DvEQFGyf/0sbrpJrluENHdPy1j/3TE=
+github.com/aws/aws-sdk-go-v2/feature/ec2/imds v1.18.20 h1:zOgq3uezl5nznfoK3ODuqbhVg1JzAGDUhXOsU0IDCAo=
+github.com/aws/aws-sdk-go-v2/feature/ec2/imds v1.18.20/go.mod h1:z/MVwUARehy6GAg/yQ1GO2IMl0k++cu1ohP9zo887wE=
+github.com/aws/aws-sdk-go-v2/internal/configsources v1.4.20 h1:CNXO7mvgThFGqOFgbNAP2nol2qAWBOGfqR/7tQlvLmc=
+github.com/aws/aws-sdk-go-v2/internal/configsources v1.4.20/go.mod h1:oydPDJKcfMhgfcgBUZaG+toBbwy8yPWubJXBVERtI4o=
+github.com/aws/aws-sdk-go-v2/internal/endpoints/v2 v2.7.20 h1:tN6W/hg+pkM+tf9XDkWUbDEjGLb+raoBMFsTodcoYKw=
+github.com/aws/aws-sdk-go-v2/internal/endpoints/v2 v2.7.20/go.mod h1:YJ898MhD067hSHA6xYCx5ts/jEd8BSOLtQDL3iZsvbc=
+github.com/aws/aws-sdk-go-v2/internal/ini v1.8.6 h1:qYQ4pzQ2Oz6WpQ8T3HvGHnZydA72MnLuFK9tJwmrbHw=
+github.com/aws/aws-sdk-go-v2/internal/ini v1.8.6/go.mod h1:O3h0IK87yXci+kg6flUKzJnWeziQUKciKrLjcatSNcY=
+github.com/aws/aws-sdk-go-v2/service/internal/accept-encoding v1.13.7 h1:5EniKhLZe4xzL7a+fU3C2tfUN4nWIqlLesfrjkuPFTY=
+github.com/aws/aws-sdk-go-v2/service/internal/accept-encoding v1.13.7/go.mod h1:x0nZssQ3qZSnIcePWLvcoFisRXJzcTVvYpAAdYX8+GI=
+github.com/aws/aws-sdk-go-v2/service/internal/presigned-url v1.13.20 h1:2HvVAIq+YqgGotK6EkMf+KIEqTISmTYh5zLpYyeTo1Y=
+github.com/aws/aws-sdk-go-v2/service/internal/presigned-url v1.13.20/go.mod h1:V4X406Y666khGa8ghKmphma/7C0DAtEQYhkq9z4vpbk=
+github.com/aws/aws-sdk-go-v2/service/signin v1.0.8 h1:0GFOLzEbOyZABS3PhYfBIx2rNBACYcKty+XGkTgw1ow=
+github.com/aws/aws-sdk-go-v2/service/signin v1.0.8/go.mod h1:LXypKvk85AROkKhOG6/YEcHFPoX+prKTowKnVdcaIxE=
+github.com/aws/aws-sdk-go-v2/service/sso v1.30.13 h1:kiIDLZ005EcKomYYITtfsjn7dtOwHDOFy7IbPXKek2o=
+github.com/aws/aws-sdk-go-v2/service/sso v1.30.13/go.mod h1:2h/xGEowcW/g38g06g3KpRWDlT+OTfxxI0o1KqayAB8=
+github.com/aws/aws-sdk-go-v2/service/ssooidc v1.35.17 h1:jzKAXIlhZhJbnYwHbvUQZEB8KfgAEuG0dc08Bkda7NU=
+github.com/aws/aws-sdk-go-v2/service/ssooidc v1.35.17/go.mod h1:Al9fFsXjv4KfbzQHGe6V4NZSZQXecFcvaIF4e70FoRA=
+github.com/aws/aws-sdk-go-v2/service/sts v1.41.9 h1:Cng+OOwCHmFljXIxpEVXAGMnBia8MSU6Ch5i9PgBkcU=
+github.com/aws/aws-sdk-go-v2/service/sts v1.41.9/go.mod h1:LrlIndBDdjA/EeXeyNBle+gyCwTlizzW5ycgWnvIxkk=
 github.com/aws/smithy-go v1.24.2 h1:FzA3bu/nt/vDvmnkg+R8Xl46gmzEDam6mZ1hzmwXFng=
 github.com/aws/smithy-go v1.24.2/go.mod h1:YE2RhdIuDbA5E5bTdciG9KrW3+TiEONeUWCqxX9i1Fc=
 github.com/aymanbagabas/go-osc52/v2 v2.0.1 h1:HwpRHbFMcZLEVr42D4p7XBqjyuxQH5SMiErDT4WkJ2k=
 github.com/aymanbagabas/go-osc52/v2 v2.0.1/go.mod h1:uYgXzlJ7ZpABp8OJ+exZzJJhRNQ2ASbcXHWsFqH8hp8=
-github.com/aymanbagabas/go-udiff v0.4.0 h1:TKnLPh7IbnizJIBKFWa9mKayRUBQ9Kh1BPCk6w2PnYM=
-github.com/aymanbagabas/go-udiff v0.4.0/go.mod h1:0L9PGwj20lrtmEMeyw4WKJ/TMyDtvAoK9bf2u/mNo3w=
 github.com/aymanbagabas/go-udiff v0.4.1 h1:OEIrQ8maEeDBXQDoGCbbTTXYJMYRCRO1fnodZ12Gv5o=
 github.com/aymanbagabas/go-udiff v0.4.1/go.mod h1:0L9PGwj20lrtmEMeyw4WKJ/TMyDtvAoK9bf2u/mNo3w=
 github.com/aymerick/douceur v0.2.0 h1:Mv+mAeH1Q+n9Fr+oyamOlAkUNPWPlA8PPGR0QAaYuPk=
 github.com/aymerick/douceur v0.2.0/go.mod h1:wlT5vV2O3h55X9m7iVYN0TBM0NH/MmbLnd30/FjWUq4=
 github.com/bahlo/generic-list-go v0.2.0 h1:5sz/EEAK+ls5wF+NeqDpk5+iNdMDXrh3z3nPnH1Wvgk=
 github.com/bahlo/generic-list-go v0.2.0/go.mod h1:2KvAjgMlE5NNynlg/5iLrrCCZ2+5xWbdbCW3pNTGyYg=
-github.com/buger/jsonparser v1.1.1 h1:2PnMjfWD7wBILjqQbt530v576A/cAbQvEW9gGIpYMUs=
-github.com/buger/jsonparser v1.1.1/go.mod h1:6RYKKt7H4d4+iWqouImQ9R2FZql3VbhNgx27UK13J/0=
-github.com/catppuccin/go v0.2.0 h1:ktBeIrIP42b/8FGiScP9sgrWOss3lw0Z5SktRoithGA=
-github.com/catppuccin/go v0.2.0/go.mod h1:8IHJuMGaUUjQM82qBrGNBv7LFq6JI3NnQCF6MOlZjpc=
+github.com/buger/jsonparser v1.1.2 h1:frqHqw7otoVbk5M8LlE/L7HTnIq2v9RX6EJ48i9AxJk=
+github.com/buger/jsonparser v1.1.2/go.mod h1:6RYKKt7H4d4+iWqouImQ9R2FZql3VbhNgx27UK13J/0=
+github.com/catppuccin/go v0.3.0 h1:d+0/YicIq+hSTo5oPuRi5kOpqkVA5tAsU6dNhvRu+aY=
+github.com/catppuccin/go v0.3.0/go.mod h1:8IHJuMGaUUjQM82qBrGNBv7LFq6JI3NnQCF6MOlZjpc=
 github.com/cespare/xxhash/v2 v2.3.0 h1:UL815xU9SqsFlibzuggzjXhog7bL6oX9BbNZnL2UFvs=
 github.com/cespare/xxhash/v2 v2.3.0/go.mod h1:VGX0DQ3Q6kWi7AoAeZDth3/j3BFtOZR5XLFGgcrjCOs=
 github.com/charmbracelet/anthropic-sdk-go v0.0.0-20260223140439-63879b0b8dab h1:J7XQLgl9sefgTnTGrmX3xqvp5o6MCiBzEjGv5igAlc4=
 github.com/charmbracelet/anthropic-sdk-go v0.0.0-20260223140439-63879b0b8dab/go.mod h1:hqlYqR7uPKOKfnNeicUbZp0Ps0GeYFlKYtwh5HGDCx8=
-github.com/charmbracelet/colorprofile v0.4.2 h1:BdSNuMjRbotnxHSfxy+PCSa4xAmz7szw70ktAtWRYrY=
-github.com/charmbracelet/colorprofile v0.4.2/go.mod h1:0rTi81QpwDElInthtrQ6Ni7cG0sDtwAd4C4le060fT8=
-github.com/charmbracelet/fang v0.4.4 h1:G4qKxF6or/eTPgmAolwPuRNyuci3hTUGGX1rj1YkHJY=
-github.com/charmbracelet/fang v0.4.4/go.mod h1:P5/DNb9DddQ0Z0dbc0P3ol4/ix5Po7Ofr2KMBfAqoCo=
-github.com/charmbracelet/glamour v0.10.0 h1:MtZvfwsYCx8jEPFJm3rIBFIMZUfUJ765oX8V6kXldcY=
-github.com/charmbracelet/glamour v0.10.0/go.mod h1:f+uf+I/ChNmqo087elLnVdCiVgjSKWuXa/l6NU2ndYk=
+github.com/charmbracelet/colorprofile v0.4.3 h1:QPa1IWkYI+AOB+fE+mg/5/4HRMZcaXex9t5KX76i20Q=
+github.com/charmbracelet/colorprofile v0.4.3/go.mod h1:/zT4BhpD5aGFpqQQqw7a+VtHCzu+zrQtt1zhMt9mR4Q=
+github.com/charmbracelet/fang v1.0.0 h1:jESBY40agJOlLYnnv9jE0mLqDGTxEk0hkOnx7YGyRlQ=
+github.com/charmbracelet/fang v1.0.0/go.mod h1:P5/DNb9DddQ0Z0dbc0P3ol4/ix5Po7Ofr2KMBfAqoCo=
+github.com/charmbracelet/glamour v1.0.0 h1:AWMLOVFHTsysl4WV8T8QgkQ0s/ZNZo7CiE4WKhk8l08=
+github.com/charmbracelet/glamour v1.0.0/go.mod h1:DSdohgOBkMr2ZQNhw4LZxSGpx3SvpeujNoXrQyH2hxo=
 github.com/charmbracelet/harmonica v0.2.0 h1:8NxJWRWg/bzKqqEaaeFNipOu77YR5t8aSwG4pgaUBiQ=
 github.com/charmbracelet/harmonica v0.2.0/go.mod h1:KSri/1RMQOZLbw7AHqgcBycp8pgJnQMYYT8QZRqZ1Ao=
 github.com/charmbracelet/lipgloss v1.1.1-0.20250404203927-76690c660834 h1:ZR7e0ro+SZZiIZD7msJyA+NjkCNNavuiPBLgerbOziE=
 github.com/charmbracelet/lipgloss v1.1.1-0.20250404203927-76690c660834/go.mod h1:aKC/t2arECF6rNOnaKaVU6y4t4ZeHQzqfxedE/VkVhA=
-github.com/charmbracelet/log v0.4.2 h1:hYt8Qj6a8yLnvR+h7MwsJv/XvmBJXiueUcI3cIxsyig=
-github.com/charmbracelet/log v0.4.2/go.mod h1:qifHGX/tc7eluv2R6pWIpyHDDrrb/AG71Pf2ysQu5nw=
-github.com/charmbracelet/ultraviolet v0.0.0-20260303162955-0b88c25f3fff h1:uY7A6hTokHPJBHfq7rj9Y/wm+IAjOghZTxKfVW6QLvw=
-github.com/charmbracelet/ultraviolet v0.0.0-20260303162955-0b88c25f3fff/go.mod h1:E6/0abq9uG2SnM8IbLB9Y5SW09uIgfaFETk8aRzgXUQ=
+github.com/charmbracelet/log v1.0.0 h1:HVVVMmfOorfj3BA9i8X8UL69Hoz9lI0PYwXfJvOdRc4=
+github.com/charmbracelet/log v1.0.0/go.mod h1:uYgY3SmLpwJWxmlrPwXvzVYujxis1vAKRV/0VQB7yWA=
+github.com/charmbracelet/openai-go v0.0.0-20260319145158-d0740cc34266 h1:BW/sZtyd1JyYy0h5adMm3tzpNyL857LWjuTRET6OhpY=
+github.com/charmbracelet/openai-go v0.0.0-20260319145158-d0740cc34266/go.mod h1:1DahUaExbUZx/jD+FNT2PKP4L9rLE5+ZBRuI8mZjd/E=
+github.com/charmbracelet/ultraviolet v0.0.0-20260316091819-b93f6a3b8502 h1:hzWNs3UQRSUTS6YCbLaQnwqKBFXT5Yh1OOw6+26apqg=
+github.com/charmbracelet/ultraviolet v0.0.0-20260316091819-b93f6a3b8502/go.mod h1:mkUCcxn9w9j89JJp3pOza5tmDQZPgIB75UfmQlFYvas=
 github.com/charmbracelet/x/ansi v0.11.6 h1:GhV21SiDz/45W9AnV2R61xZMRri5NlLnl6CVF7ihZW8=
 github.com/charmbracelet/x/ansi v0.11.6/go.mod h1:2JNYLgQUsyqaiLovhU2Rv/pb8r6ydXKS3NIttu3VGZQ=
 github.com/charmbracelet/x/cellbuf v0.0.15 h1:ur3pZy0o6z/R7EylET877CBxaiE1Sp1GMxoFPAIztPI=
 github.com/charmbracelet/x/cellbuf v0.0.15/go.mod h1:J1YVbR7MUuEGIFPCaaZ96KDl5NoS0DAWkskup+mOY+Q=
-github.com/charmbracelet/x/exp/charmtone v0.0.0-20260305213658-fe36e8c10185 h1:/192monmpmRICpSPrFRzkIO+xfhioV6/nwrQdkDTj10=
-github.com/charmbracelet/x/exp/charmtone v0.0.0-20260305213658-fe36e8c10185/go.mod h1:nsExn0DGyX0lh9LwLHTn2Gg+hafdzfSXnC+QmEJTZFY=
+github.com/charmbracelet/x/conpty v0.1.1 h1:s1bUxjoi7EpqiXysVtC+a8RrvPPNcNvAjfi4jxsAuEs=
+github.com/charmbracelet/x/conpty v0.1.1/go.mod h1:OmtR77VODEFbiTzGE9G1XiRJAga6011PIm4u5fTNZpk=
+github.com/charmbracelet/x/errors v0.0.0-20240508181413-e8d8b6e2de86 h1:JSt3B+U9iqk37QUU2Rvb6DSBYRLtWqFqfxf8l5hOZUA=
+github.com/charmbracelet/x/errors v0.0.0-20240508181413-e8d8b6e2de86/go.mod h1:2P0UgXMEa6TsToMSuFqKFQR+fZTO9CNGUNokkPatT/0=
+github.com/charmbracelet/x/exp/charmtone v0.0.0-20260322003602-9b007323c5cd h1:eStB6uX52pgrm6TxQcEKctPrEC+a/9ubJC+P671idOc=
+github.com/charmbracelet/x/exp/charmtone v0.0.0-20260322003602-9b007323c5cd/go.mod h1:nsExn0DGyX0lh9LwLHTn2Gg+hafdzfSXnC+QmEJTZFY=
 github.com/charmbracelet/x/exp/golden v0.0.0-20250806222409-83e3a29d542f h1:pk6gmGpCE7F3FcjaOEKYriCvpmIN4+6OS/RD0vm4uIA=
 github.com/charmbracelet/x/exp/golden v0.0.0-20250806222409-83e3a29d542f/go.mod h1:IfZAMTHB6XkZSeXUqriemErjAWCCzT0LwjKFYCZyw0I=
 github.com/charmbracelet/x/exp/ordered v0.1.0 h1:55/qLwjIh0gL0Vni+QAWk7T/qRVP6sBf+2agPBgnOFE=
 github.com/charmbracelet/x/exp/ordered v0.1.0/go.mod h1:5UHwmG+is5THxMyCJHNPCn2/ecI07aKNrW+LcResjJ8=
-github.com/charmbracelet/x/exp/slice v0.0.0-20260305213658-fe36e8c10185 h1:bloHJLweYZeIkBVgi8AF94DrTdx3eoEB57VOpFuFi3U=
-github.com/charmbracelet/x/exp/slice v0.0.0-20260305213658-fe36e8c10185/go.mod h1:vqEfX6xzqW1pKKZUUiFOKg0OQ7bCh54Q2vR/tserrRA=
-github.com/charmbracelet/x/exp/strings v0.0.0-20240722160745-212f7b056ed0 h1:qko3AQ4gK1MTS/de7F5hPGx6/k1u0w4TeYmBFwzYVP4=
-github.com/charmbracelet/x/exp/strings v0.0.0-20240722160745-212f7b056ed0/go.mod h1:pBhA0ybfXv6hDjQUZ7hk1lVxBiUbupdw5R31yPUViVQ=
+github.com/charmbracelet/x/exp/slice v0.0.0-20260322003602-9b007323c5cd h1:U8xj0UXwqHzO+UYHZJopKF+gWaQEW8oj60fmiq9TFY4=
+github.com/charmbracelet/x/exp/slice v0.0.0-20260322003602-9b007323c5cd/go.mod h1:vqEfX6xzqW1pKKZUUiFOKg0OQ7bCh54Q2vR/tserrRA=
+github.com/charmbracelet/x/exp/strings v0.1.0 h1:i69S2XI7uG1u4NLGeJPSYU++Nmjvpo9nwd6aoEm7gkA=
+github.com/charmbracelet/x/exp/strings v0.1.0/go.mod h1:/ehtMPNh9K4odGFkqYJKpIYyePhdp1hLBRvyY4bWkH8=
 github.com/charmbracelet/x/json v0.2.0 h1:DqB+ZGx2h+Z+1s98HOuOyli+i97wsFQIxP2ZQANTPrQ=
 github.com/charmbracelet/x/json v0.2.0/go.mod h1:opFIflx2YgXgi49xVUu8gEQ21teFAxyMwvOiZhIvWNM=
 github.com/charmbracelet/x/term v0.2.2 h1:xVRT/S2ZcKdhhOuSP4t5cLi5o+JxklsoEObBSgfgZRk=
@@ -122,6 +122,8 @@ github.com/charmbracelet/x/termios v0.1.1 h1:o3Q2bT8eqzGnGPOYheoYS8eEleT5ZVNYNy8
 github.com/charmbracelet/x/termios v0.1.1/go.mod h1:rB7fnv1TgOPOyyKRJ9o+AsTU/vK5WHJ2ivHeut/Pcwo=
 github.com/charmbracelet/x/windows v0.2.2 h1:IofanmuvaxnKHuV04sC0eBy/smG6kIKrWG2/jYn2GuM=
 github.com/charmbracelet/x/windows v0.2.2/go.mod h1:/8XtdKZzedat74NQFn0NGlGL4soHB0YQZrETF96h75k=
+github.com/charmbracelet/x/xpty v0.1.3 h1:eGSitii4suhzrISYH50ZfufV3v085BXQwIytcOdFSsw=
+github.com/charmbracelet/x/xpty v0.1.3/go.mod h1:poPYpWuLDBFCKmKLDnhBp51ATa0ooD8FhypRwEFtH3Y=
 github.com/clipperhouse/displaywidth v0.11.0 h1:lBc6kY44VFw+TDx4I8opi/EtL9m20WSEFgwIwO+UVM8=
 github.com/clipperhouse/displaywidth v0.11.0/go.mod h1:bkrFNkf81G8HyVqmKGxsPufD3JhNl3dSqnGhOoSD/o0=
 github.com/clipperhouse/uax29/v2 v2.7.0 h1:+gs4oBZ2gPfVrKPthwbMzWZDaAFPGYK72F0NJv2v7Vk=
@@ -131,6 +133,8 @@ github.com/cncf/xds/go v0.0.0-20260202195803-dba9d589def2/go.mod h1:qwXFYgsP6T7X
 github.com/coder/acp-go-sdk v0.6.3 h1:LsXQytehdjKIYJnoVWON/nf7mqbiarnyuyE3rrjBsXQ=
 github.com/coder/acp-go-sdk v0.6.3/go.mod h1:yKzM/3R9uELp4+nBAwwtkS0aN1FOFjo11CNPy37yFko=
 github.com/cpuguy83/go-md2man/v2 v2.0.6/go.mod h1:oOW0eioCTA6cOiMLiUPZOpcVxMig6NIQQ7OS05n1F4g=
+github.com/creack/pty v1.1.24 h1:bJrF4RRfyJnbTJqzRLHzcGaZK1NeM5kTC9jGgovnR1s=
+github.com/creack/pty v1.1.24/go.mod h1:08sCNb52WyoAwi2QDyzUCTgcvVFhUzewun7wtTfvcwE=
 github.com/davecgh/go-spew v1.1.2-0.20180830191138-d8f796af33cc h1:U9qPSI2PIWSS1VwoXQT9A3Wy9MM3WgvqSxFWenqJduM=
 github.com/davecgh/go-spew v1.1.2-0.20180830191138-d8f796af33cc/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/dlclark/regexp2 v1.11.5 h1:Q/sSnsKerHeCkc/jSTNq1oCm7KiVgUMZRDUoRu0JQZQ=
@@ -175,8 +179,8 @@ github.com/google/uuid v1.6.0 h1:NIvaJDMOsjHA8n1jAhLSgzrAzy1Hgr+hNrb57e+94F0=
 github.com/google/uuid v1.6.0/go.mod h1:TIyPZe4MgqvfeYDBFedMoGGpEw/LqOeaOT+nhxU+yHo=
 github.com/googleapis/enterprise-certificate-proxy v0.3.14 h1:yh8ncqsbUY4shRD5dA6RlzjJaT4hi3kII+zYw8wmLb8=
 github.com/googleapis/enterprise-certificate-proxy v0.3.14/go.mod h1:vqVt9yG9480NtzREnTlmGSBmFrA+bzb0yl0TxoBQXOg=
-github.com/googleapis/gax-go/v2 v2.17.0 h1:RksgfBpxqff0EZkDWYuz9q/uWsTVz+kf43LsZ1J6SMc=
-github.com/googleapis/gax-go/v2 v2.17.0/go.mod h1:mzaqghpQp4JDh3HvADwrat+6M3MOIDp5YKHhb9PAgDY=
+github.com/googleapis/gax-go/v2 v2.19.0 h1:fYQaUOiGwll0cGj7jmHT/0nPlcrZDFPrZRhTsoCr8hE=
+github.com/googleapis/gax-go/v2 v2.19.0/go.mod h1:w2ROXVdfGEVFXzmlciUU4EdjHgWvB5h2n6x/8XSTTJA=
 github.com/gorilla/css v1.0.1 h1:ntNaBIghp6JmvWnxbZKANoLyuXTPZ4cAMlo6RyhlbO8=
 github.com/gorilla/css v1.0.1/go.mod h1:BvnYkspnSzMmwRK+b8/xgNPLiIuNZr6vbZBTPQ2A3b0=
 github.com/gorilla/websocket v1.5.3 h1:saDtZ6Pbx/0u+bgYQ3q96pZgCzfhKXGPqt7kZ72aNNg=
@@ -191,8 +195,8 @@ github.com/kaptinlin/go-i18n v0.2.12 h1:ywDsvb4KDFddMC2dpI/rrIzGU2mWUSvHmWUm9BMs
 github.com/kaptinlin/go-i18n v0.2.12/go.mod h1:pVcu9qsW5pOIOoZFJXesRYmLos1vMQrby70JPAoWmJU=
 github.com/kaptinlin/jsonpointer v0.4.17 h1:mY9k8ciWncxbsECyaxKnR0MdmxamNdp2tLQkAKVrtSk=
 github.com/kaptinlin/jsonpointer v0.4.17/go.mod h1:SsfsjqnHG5zuKo1DTBzk1VknaHlL4osHw+X9kZKukpU=
-github.com/kaptinlin/jsonschema v0.7.5 h1:jkK4a3NyzNoGlvu12CsL3IcqNMVa5sL51HPVa0nWcPY=
-github.com/kaptinlin/jsonschema v0.7.5/go.mod h1:3gIWnptl+SWMyfMR2r4TXXd0xsQZ1m50AKrwmcUONSg=
+github.com/kaptinlin/jsonschema v0.7.6 h1:UUMqZGFAk7nOzQsYAxvgygm4wpDp/nwXxA4VP9mCPCs=
+github.com/kaptinlin/jsonschema v0.7.6/go.mod h1:GGk/oE+F1lWUfYrzKaCf4QWZmMdytt0LL4XdFEFB0LE=
 github.com/kaptinlin/messageformat-go v0.4.18 h1:RBlHVWgZyoxTcUgGWBsl2AcyScq/urqbLZvzgryTmSI=
 github.com/kaptinlin/messageformat-go v0.4.18/go.mod h1:ntI3154RnqJgr7GaC+vZBnIExl2V3sv9selvRNNEM24=
 github.com/kr/pretty v0.3.1 h1:flRD4NNwYAUpkphVc1HcthR4KEIFJ65n8Mw5qdRn3LE=
@@ -203,15 +207,15 @@ github.com/kylelemons/godebug v1.1.0 h1:RPNrshWIDI6G2gRW9EHilWtl7Z6Sb1BR0xunSBf0
 github.com/kylelemons/godebug v1.1.0/go.mod h1:9/0rRGxNHcop5bhtWyNeEfOS8JIWk580+fNqagV/RAw=
 github.com/lucasb-eyer/go-colorful v1.3.0 h1:2/yBRLdWBZKrf7gB40FoiKfAWYQ0lqNcbuQwVHXptag=
 github.com/lucasb-eyer/go-colorful v1.3.0/go.mod h1:R4dSotOR9KMtayYi1e77YzuveK+i7ruzyGqttikkLy0=
-github.com/mailru/easyjson v0.9.1 h1:LbtsOm5WAswyWbvTEOqhypdPeZzHavpZx96/n553mR8=
-github.com/mailru/easyjson v0.9.1/go.mod h1:1+xMtQp2MRNVL/V1bOzuP3aP8VNwRW55fQUto+XFtTU=
-github.com/mark3labs/mcp-go v0.44.1 h1:2PKppYlT9X2fXnE8SNYQLAX4hNjfPB0oNLqQVcN6mE8=
-github.com/mark3labs/mcp-go v0.44.1/go.mod h1:YnJfOL382MIWDx1kMY+2zsRHU/q78dBg9aFb8W6Thdw=
+github.com/mailru/easyjson v0.9.2 h1:dX8U45hQsZpxd80nLvDGihsQ/OxlvTkVUXH2r/8cb2M=
+github.com/mailru/easyjson v0.9.2/go.mod h1:1+xMtQp2MRNVL/V1bOzuP3aP8VNwRW55fQUto+XFtTU=
+github.com/mark3labs/mcp-go v0.45.0 h1:s0S8qR/9fWaQ3pHxz7pm1uQ0DrswoSnRIxKIjbiQtkc=
+github.com/mark3labs/mcp-go v0.45.0/go.mod h1:YnJfOL382MIWDx1kMY+2zsRHU/q78dBg9aFb8W6Thdw=
 github.com/mattn/go-isatty v0.0.20 h1:xfD0iDuEKnDkl03q4limB+vH+GxLEtL/jb4xVJSWWEY=
 github.com/mattn/go-isatty v0.0.20/go.mod h1:W+V8PltTTMOvKvAeJH7IuucS94S2C6jfK/D7dTCTo3Y=
 github.com/mattn/go-runewidth v0.0.12/go.mod h1:RAqKPSqVFrSLVXbA8x7dzmKdmGzieGRCM46jaSJTDAk=
-github.com/mattn/go-runewidth v0.0.20 h1:WcT52H91ZUAwy8+HUkdM3THM6gXqXuLJi9O3rjcQQaQ=
-github.com/mattn/go-runewidth v0.0.20/go.mod h1:XBkDxAl56ILZc9knddidhrOlY5R/pDhgLpndooCuJAs=
+github.com/mattn/go-runewidth v0.0.21 h1:jJKAZiQH+2mIinzCJIaIG9Be1+0NR+5sz/lYEEjdM8w=
+github.com/mattn/go-runewidth v0.0.21/go.mod h1:XBkDxAl56ILZc9knddidhrOlY5R/pDhgLpndooCuJAs=
 github.com/microcosm-cc/bluemonday v1.0.27 h1:MpEUotklkwCSLeH+Qdx1VJgNqLlpY2KXwXFM08ygZfk=
 github.com/microcosm-cc/bluemonday v1.0.27/go.mod h1:jFi9vgW+H7c3V0lb6nR74Ib/DIB5OBs92Dimizgw2cA=
 github.com/mitchellh/hashstructure/v2 v2.0.2 h1:vGKWl0YJqUNxE8d+h8f6NJLcCJrgbhC4NcD46KavDd4=
@@ -230,8 +234,6 @@ github.com/muesli/roff v0.1.0 h1:YD0lalCotmYuF5HhZliKWlIx7IEhiXeSfq7hNjFqGF8=
 github.com/muesli/roff v0.1.0/go.mod h1:pjAHQM9hdUUwm/krAfrLGgJkXJ+YuhtsfZ42kieB2Ig=
 github.com/muesli/termenv v0.16.0 h1:S5AlUN9dENB57rsbnkPyfdGuWIlkmzJjbFf0Tf5FWUc=
 github.com/muesli/termenv v0.16.0/go.mod h1:ZRfOIKPFDYQoDFF4Olj7/QJbW60Ol/kL1pU3VfY/Cnk=
-github.com/openai/openai-go/v2 v2.7.1 h1:/tfvTJhfv7hTSL8mWwc5VL4WLLSDL5yn9VqVykdu9r8=
-github.com/openai/openai-go/v2 v2.7.1/go.mod h1:jrJs23apqJKKbT+pqtFgNKpRju/KP9zpUTZhz3GElQE=
 github.com/pelletier/go-toml/v2 v2.2.4 h1:mye9XuhQ6gvn5h28+VilKrrPoQVanw5PMw/TB0t5Ec4=
 github.com/pelletier/go-toml/v2 v2.2.4/go.mod h1:2gIqNv+qfxSVS7cM2xJQKtLSTLUE9V8t9Stt+h56mCY=
 github.com/pkg/browser v0.0.0-20240102092130-5ac0b6a4141c h1:+mdjkGKdHQG3305AYmdv1U2eRNDiU2ErMBj1gwrq8eQ=
@@ -283,59 +285,57 @@ github.com/xo/terminfo v0.0.0-20220910002029-abceb7e1c41e h1:JVG44RsyaB9T2KIHavM
 github.com/xo/terminfo v0.0.0-20220910002029-abceb7e1c41e/go.mod h1:RbqR21r5mrJuqunuUZ/Dhy/avygyECGrLceyNeo4LiM=
 github.com/yosida95/uritemplate/v3 v3.0.2 h1:Ed3Oyj9yrmi9087+NczuL5BwkIc4wvTb5zIM+UJPGz4=
 github.com/yosida95/uritemplate/v3 v3.0.2/go.mod h1:ILOh0sOhIJR3+L/8afwt/kE++YT040gmv5BQTMR2HP4=
-github.com/yuin/goldmark v1.7.16 h1:n+CJdUxaFMiDUNnWC3dMWCIQJSkxH4uz3ZwQBkAlVNE=
-github.com/yuin/goldmark v1.7.16/go.mod h1:ip/1k0VRfGynBgxOz0yCqHrbZXhcjxyuS66Brc7iBKg=
+github.com/yuin/goldmark v1.7.17 h1:p36OVWwRb246iHxA/U4p8OPEpOTESm4n+g+8t0EE5uA=
+github.com/yuin/goldmark v1.7.17/go.mod h1:ip/1k0VRfGynBgxOz0yCqHrbZXhcjxyuS66Brc7iBKg=
 github.com/yuin/goldmark-emoji v1.0.6 h1:QWfF2FYaXwL74tfGOW5izeiZepUDroDJfWubQI9HTHs=
 github.com/yuin/goldmark-emoji v1.0.6/go.mod h1:ukxJDKFpdFb5x0a5HqbdlcKtebh086iJpI31LTKmWuA=
 go.opentelemetry.io/auto/sdk v1.2.1 h1:jXsnJ4Lmnqd11kwkBV2LgLoFMZKizbCi5fNZ/ipaZ64=
 go.opentelemetry.io/auto/sdk v1.2.1/go.mod h1:KRTj+aOaElaLi+wW1kO/DZRXwkF4C5xPbEe3ZiIhN7Y=
-go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc v0.66.0 h1:w/o339tDd6Qtu3+ytwt+/jon2yjAs3Ot8Xq8pelfhSo=
-go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc v0.66.0/go.mod h1:pdhNtM9C4H5fRdrnwO7NjxzQWhKSSxCHk/KluVqDVC0=
-go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.66.0 h1:PnV4kVnw0zOmwwFkAzCN5O07fw1YOIQor120zrh0AVo=
-go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.66.0/go.mod h1:ofAwF4uinaf8SXdVzzbL4OsxJ3VfeEg3f/F6CeF49/Y=
-go.opentelemetry.io/otel v1.41.0 h1:YlEwVsGAlCvczDILpUXpIpPSL/VPugt7zHThEMLce1c=
-go.opentelemetry.io/otel v1.41.0/go.mod h1:Yt4UwgEKeT05QbLwbyHXEwhnjxNO6D8L5PQP51/46dE=
-go.opentelemetry.io/otel/metric v1.41.0 h1:rFnDcs4gRzBcsO9tS8LCpgR0dxg4aaxWlJxCno7JlTQ=
-go.opentelemetry.io/otel/metric v1.41.0/go.mod h1:xPvCwd9pU0VN8tPZYzDZV/BMj9CM9vs00GuBjeKhJps=
-go.opentelemetry.io/otel/sdk v1.41.0 h1:YPIEXKmiAwkGl3Gu1huk1aYWwtpRLeskpV+wPisxBp8=
-go.opentelemetry.io/otel/sdk v1.41.0/go.mod h1:ahFdU0G5y8IxglBf0QBJXgSe7agzjE4GiTJ6HT9ud90=
-go.opentelemetry.io/otel/sdk/metric v1.41.0 h1:siZQIYBAUd1rlIWQT2uCxWJxcCO7q3TriaMlf08rXw8=
-go.opentelemetry.io/otel/sdk/metric v1.41.0/go.mod h1:HNBuSvT7ROaGtGI50ArdRLUnvRTRGniSUZbxiWxSO8Y=
-go.opentelemetry.io/otel/trace v1.41.0 h1:Vbk2co6bhj8L59ZJ6/xFTskY+tGAbOnCtQGVVa9TIN0=
-go.opentelemetry.io/otel/trace v1.41.0/go.mod h1:U1NU4ULCoxeDKc09yCWdWe+3QoyweJcISEVa1RBzOis=
+go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc v0.67.0 h1:yI1/OhfEPy7J9eoa6Sj051C7n5dvpj0QX8g4sRchg04=
+go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc v0.67.0/go.mod h1:NoUCKYWK+3ecatC4HjkRktREheMeEtrXoQxrqYFeHSc=
+go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.67.0 h1:OyrsyzuttWTSur2qN/Lm0m2a8yqyIjUVBZcxFPuXq2o=
+go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.67.0/go.mod h1:C2NGBr+kAB4bk3xtMXfZ94gqFDtg/GkI7e9zqGh5Beg=
+go.opentelemetry.io/otel v1.42.0 h1:lSQGzTgVR3+sgJDAU/7/ZMjN9Z+vUip7leaqBKy4sho=
+go.opentelemetry.io/otel v1.42.0/go.mod h1:lJNsdRMxCUIWuMlVJWzecSMuNjE7dOYyWlqOXWkdqCc=
+go.opentelemetry.io/otel/metric v1.42.0 h1:2jXG+3oZLNXEPfNmnpxKDeZsFI5o4J+nz6xUlaFdF/4=
+go.opentelemetry.io/otel/metric v1.42.0/go.mod h1:RlUN/7vTU7Ao/diDkEpQpnz3/92J9ko05BIwxYa2SSI=
+go.opentelemetry.io/otel/sdk v1.42.0 h1:LyC8+jqk6UJwdrI/8VydAq/hvkFKNHZVIWuslJXYsDo=
+go.opentelemetry.io/otel/sdk v1.42.0/go.mod h1:rGHCAxd9DAph0joO4W6OPwxjNTYWghRWmkHuGbayMts=
+go.opentelemetry.io/otel/sdk/metric v1.42.0 h1:D/1QR46Clz6ajyZ3G8SgNlTJKBdGp84q9RKCAZ3YGuA=
+go.opentelemetry.io/otel/sdk/metric v1.42.0/go.mod h1:Ua6AAlDKdZ7tdvaQKfSmnFTdHx37+J4ba8MwVCYM5hc=
+go.opentelemetry.io/otel/trace v1.42.0 h1:OUCgIPt+mzOnaUTpOQcBiM/PLQ/Op7oq6g4LenLmOYY=
+go.opentelemetry.io/otel/trace v1.42.0/go.mod h1:f3K9S+IFqnumBkKhRJMeaZeNk9epyhnCmQh/EysQCdc=
 go.yaml.in/yaml/v3 v3.0.4 h1:tfq32ie2Jv2UxXFdLJdh3jXuOzWiL1fo0bu/FbuKpbc=
 go.yaml.in/yaml/v3 v3.0.4/go.mod h1:DhzuOOF2ATzADvBadXxruRBLzYTpT36CKvDb3+aBEFg=
-golang.org/x/crypto v0.48.0 h1:/VRzVqiRSggnhY7gNRxPauEQ5Drw9haKdM0jqfcCFts=
-golang.org/x/crypto v0.48.0/go.mod h1:r0kV5h3qnFPlQnBSrULhlsRfryS2pmewsg+XfMgkVos=
-golang.org/x/exp v0.0.0-20260218203240-3dfff04db8fa h1:Zt3DZoOFFYkKhDT3v7Lm9FDMEV06GpzjG2jrqW+QTE0=
-golang.org/x/exp v0.0.0-20260218203240-3dfff04db8fa/go.mod h1:K79w1Vqn7PoiZn+TkNpx3BUWUQksGO3JcVX6qIjytmA=
-golang.org/x/net v0.51.0 h1:94R/GTO7mt3/4wIKpcR5gkGmRLOuE/2hNGeWq/GBIFo=
-golang.org/x/net v0.51.0/go.mod h1:aamm+2QF5ogm02fjy5Bb7CQ0WMt1/WVM7FtyaTLlA9Y=
-golang.org/x/oauth2 v0.35.0 h1:Mv2mzuHuZuY2+bkyWXIHMfhNdJAdwW3FuWeCPYN5GVQ=
-golang.org/x/oauth2 v0.35.0/go.mod h1:lzm5WQJQwKZ3nwavOZ3IS5Aulzxi68dUSgRHujetwEA=
-golang.org/x/sync v0.19.0 h1:vV+1eWNmZ5geRlYjzm2adRgW2/mcpevXNg50YZtPCE4=
-golang.org/x/sync v0.19.0/go.mod h1:9KTHXmSnoGruLpwFjVSX0lNNA75CykiMECbovNTZqGI=
+golang.org/x/crypto v0.49.0 h1:+Ng2ULVvLHnJ/ZFEq4KdcDd/cfjrrjjNSXNzxg0Y4U4=
+golang.org/x/crypto v0.49.0/go.mod h1:ErX4dUh2UM+CFYiXZRTcMpEcN8b/1gxEuv3nODoYtCA=
+golang.org/x/exp v0.0.0-20260312153236-7ab1446f8b90 h1:jiDhWWeC7jfWqR9c/uplMOqJ0sbNlNWv0UkzE0vX1MA=
+golang.org/x/exp v0.0.0-20260312153236-7ab1446f8b90/go.mod h1:xE1HEv6b+1SCZ5/uscMRjUBKtIxworgEcEi+/n9NQDQ=
+golang.org/x/net v0.52.0 h1:He/TN1l0e4mmR3QqHMT2Xab3Aj3L9qjbhRm78/6jrW0=
+golang.org/x/net v0.52.0/go.mod h1:R1MAz7uMZxVMualyPXb+VaqGSa3LIaUqk0eEt3w36Sw=
+golang.org/x/oauth2 v0.36.0 h1:peZ/1z27fi9hUOFCAZaHyrpWG5lwe0RJEEEeH0ThlIs=
+golang.org/x/oauth2 v0.36.0/go.mod h1:YDBUJMTkDnJS+A4BP4eZBjCqtokkg1hODuPjwiGPO7Q=
+golang.org/x/sync v0.20.0 h1:e0PTpb7pjO8GAtTs2dQ6jYa5BWYlMuX047Dco/pItO4=
+golang.org/x/sync v0.20.0/go.mod h1:9xrNwdLfx4jkKbNva9FpL6vEN7evnE43NNNJQ2LF3+0=
 golang.org/x/sys v0.6.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
-golang.org/x/sys v0.41.0 h1:Ivj+2Cp/ylzLiEU89QhWblYnOE9zerudt9Ftecq2C6k=
-golang.org/x/sys v0.41.0/go.mod h1:OgkHotnGiDImocRcuBABYBEXf8A9a87e/uXjp9XT3ks=
 golang.org/x/sys v0.42.0 h1:omrd2nAlyT5ESRdCLYdm3+fMfNFE/+Rf4bDIQImRJeo=
 golang.org/x/sys v0.42.0/go.mod h1:4GL1E5IUh+htKOUEOaiffhrAeqysfVGipDYzABqnCmw=
-golang.org/x/term v0.40.0 h1:36e4zGLqU4yhjlmxEaagx2KuYbJq3EwY8K943ZsHcvg=
-golang.org/x/term v0.40.0/go.mod h1:w2P8uVp06p2iyKKuvXIm7N/y0UCRt3UfJTfZ7oOpglM=
-golang.org/x/text v0.34.0 h1:oL/Qq0Kdaqxa1KbNeMKwQq0reLCCaFtqu2eNuSeNHbk=
-golang.org/x/text v0.34.0/go.mod h1:homfLqTYRFyVYemLBFl5GgL/DWEiH5wcsQ5gSh1yziA=
-golang.org/x/time v0.14.0 h1:MRx4UaLrDotUKUdCIqzPC48t1Y9hANFKIRpNx+Te8PI=
-golang.org/x/time v0.14.0/go.mod h1:eL/Oa2bBBK0TkX57Fyni+NgnyQQN4LitPmob2Hjnqw4=
+golang.org/x/term v0.41.0 h1:QCgPso/Q3RTJx2Th4bDLqML4W6iJiaXFq2/ftQF13YU=
+golang.org/x/term v0.41.0/go.mod h1:3pfBgksrReYfZ5lvYM0kSO0LIkAl4Yl2bXOkKP7Ec2A=
+golang.org/x/text v0.35.0 h1:JOVx6vVDFokkpaq1AEptVzLTpDe9KGpj5tR4/X+ybL8=
+golang.org/x/text v0.35.0/go.mod h1:khi/HExzZJ2pGnjenulevKNX1W67CUy0AsXcNubPGCA=
+golang.org/x/time v0.15.0 h1:bbrp8t3bGUeFOx08pvsMYRTCVSMk89u4tKbNOZbp88U=
+golang.org/x/time v0.15.0/go.mod h1:Y4YMaQmXwGQZoFaVFk4YpCt4FLQMYKZe9oeV/f4MSno=
 gonum.org/v1/gonum v0.16.0 h1:5+ul4Swaf3ESvrOnidPp4GZbzf0mxVQpDCYUQE7OJfk=
 gonum.org/v1/gonum v0.16.0/go.mod h1:fef3am4MQ93R2HHpKnLk4/Tbh/s0+wqD5nfa6Pnwy4E=
-google.golang.org/api v0.269.0 h1:qDrTOxKUQ/P0MveH6a7vZ+DNHxJQjtGm/uvdbdGXCQg=
-google.golang.org/api v0.269.0/go.mod h1:N8Wpcu23Tlccl0zSHEkcAZQKDLdquxK+l9r2LkwAauE=
-google.golang.org/genai v1.49.0 h1:Se+QJaH2GYK1aaR1o5S38mlU2GD5FnVvP76nfkV7LH0=
-google.golang.org/genai v1.49.0/go.mod h1:A3kkl0nyBjyFlNjgxIwKq70julKbIxpSxqKO5gw/gmk=
-google.golang.org/genproto/googleapis/rpc v0.0.0-20260226221140-a57be14db171 h1:ggcbiqK8WWh6l1dnltU4BgWGIGo+EVYxCaAPih/zQXQ=
-google.golang.org/genproto/googleapis/rpc v0.0.0-20260226221140-a57be14db171/go.mod h1:4Hqkh8ycfw05ld/3BWL7rJOSfebL2Q+DVDeRgYgxUU8=
-google.golang.org/grpc v1.79.2 h1:fRMD94s2tITpyJGtBBn7MkMseNpOZU8ZxgC3MMBaXRU=
-google.golang.org/grpc v1.79.2/go.mod h1:KmT0Kjez+0dde/v2j9vzwoAScgEPx/Bw1CYChhHLrHQ=
+google.golang.org/api v0.272.0 h1:eLUQZGnAS3OHn31URRf9sAmRk3w2JjMx37d2k8AjJmA=
+google.golang.org/api v0.272.0/go.mod h1:wKjowi5LNJc5qarNvDCvNQBn3rVK8nSy6jg2SwRwzIA=
+google.golang.org/genai v1.51.0 h1:IZGuUqgfx40INv3hLFGCbOSGp0qFqm7LVmDghzNIYqg=
+google.golang.org/genai v1.51.0/go.mod h1:A3kkl0nyBjyFlNjgxIwKq70julKbIxpSxqKO5gw/gmk=
+google.golang.org/genproto/googleapis/rpc v0.0.0-20260319201613-d00831a3d3e7 h1:ndE4FoJqsIceKP2oYSnUZqhTdYufCYYkqwtFzfrhI7w=
+google.golang.org/genproto/googleapis/rpc v0.0.0-20260319201613-d00831a3d3e7/go.mod h1:4Hqkh8ycfw05ld/3BWL7rJOSfebL2Q+DVDeRgYgxUU8=
+google.golang.org/grpc v1.79.3 h1:sybAEdRIEtvcD68Gx7dmnwjZKlyfuc61Dyo9pGXXkKE=
+google.golang.org/grpc v1.79.3/go.mod h1:KmT0Kjez+0dde/v2j9vzwoAScgEPx/Bw1CYChhHLrHQ=
 google.golang.org/protobuf v1.36.11 h1:fV6ZwhNocDyBLK0dj+fg8ektcVegBBuEolpbTQyBNVE=
 google.golang.org/protobuf v1.36.11/go.mod h1:HTf+CrKn2C3g5S8VImy6tdcUvCska2kB7j23XfzDpco=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=

internal/agent/agent.go

@@ -63,6 +63,13 @@ type ToolCallContentHandler func(content string)
 // ReasoningDeltaHandler is a function type for handling streaming reasoning/thinking deltas.
 type ReasoningDeltaHandler func(delta string)
 
+// ToolOutputHandler is a function type for handling streaming tool output chunks.
+// Used by tools like bash to stream output as it arrives rather than waiting
+// for the command to complete. The isStderr flag indicates if the chunk
+// contains stderr output.
+// Note: This is an alias for core.ToolOutputCallback to avoid import cycles.
+type ToolOutputHandler = core.ToolOutputCallback
+
 // Agent represents an AI agent with core tool integration using the fantasy library.
 // Core tools (bash, read, write, edit, grep, find, ls) are registered as direct
 // fantasy.AgentTool implementations — no MCP layer, no serialization overhead.
@@ -218,7 +225,7 @@ func (a *Agent) GenerateWithLoop(ctx context.Context, messages []fantasy.Message
 	onResponse ResponseHandler, onToolCallContent ToolCallContentHandler,
 ) (*GenerateWithLoopResult, error) {
 	return a.GenerateWithLoopAndStreaming(ctx, messages, onToolCall, onToolExecution, onToolResult,
-		onResponse, onToolCallContent, nil, nil)
+		onResponse, onToolCallContent, nil, nil, nil)
 }
 
 // GenerateWithLoopAndStreaming processes messages using the fantasy agent with streaming and callbacks.
@@ -229,8 +236,14 @@ func (a *Agent) GenerateWithLoopAndStreaming(ctx context.Context, messages []fan
 	onResponse ResponseHandler, onToolCallContent ToolCallContentHandler,
 	onStreamingResponse StreamingResponseHandler,
 	onReasoningDelta ReasoningDeltaHandler,
+	onToolOutput ToolOutputHandler,
 ) (*GenerateWithLoopResult, error) {
 
+	// Inject tool output handler into context for use by core tools (e.g., bash).
+	if onToolOutput != nil {
+		ctx = core.ContextWithToolOutputCallback(ctx, onToolOutput)
+	}
+
 	// Fantasy requires the current user input as Prompt, with prior messages as history.
 	// Extract the last user message text and files as the prompt, and pass everything
 	// before it as Messages. Files (e.g. clipboard images) are passed via the Files
@@ -249,6 +262,12 @@ func (a *Agent) GenerateWithLoopAndStreaming(ctx context.Context, messages []fan
 		onToolCallContent != nil || onStreamingResponse != nil || onReasoningDelta != nil
 
 	if a.streamingEnabled || hasCallbacks {
+		// Track completed step messages so we can return partial results
+		// on cancellation. Fantasy's Stream() discards accumulated steps
+		// when it returns an error, but the OnStepFinish callback fires
+		// for every step that completed before the error occurred.
+		var completedStepMessages []fantasy.Message
+
 		// Use fantasy's streaming agent
 		result, err := a.fantasyAgent.Stream(ctx, fantasy.AgentStreamCall{
 			Prompt:   prompt,
@@ -319,6 +338,10 @@ func (a *Agent) GenerateWithLoopAndStreaming(ctx context.Context, messages []fan
 
 			// Step callbacks for content that accompanies tool calls
 			OnStepFinish: func(step fantasy.StepResult) error {
+				// Accumulate messages from completed steps so they can be
+				// persisted even if a later step is cancelled.
+				completedStepMessages = append(completedStepMessages, step.Messages...)
+
 				if ctx.Err() != nil {
 					return ctx.Err()
 				}
@@ -332,6 +355,20 @@ func (a *Agent) GenerateWithLoopAndStreaming(ctx context.Context, messages []fan
 			},
 		})
 		if err != nil {
+			// On cancellation (or any error), return a partial result
+			// containing messages from completed steps so the caller can
+			// persist tool calls and results that finished before the
+			// cancellation. The original input messages are included so
+			// the caller sees the full conversation up to the point of
+			// cancellation.
+			if len(completedStepMessages) > 0 {
+				partialMessages := make([]fantasy.Message, 0, len(messages)+len(completedStepMessages))
+				partialMessages = append(partialMessages, messages...)
+				partialMessages = append(partialMessages, completedStepMessages...)
+				return &GenerateWithLoopResult{
+					ConversationMessages: partialMessages,
+				}, err
+			}
 			return nil, err
 		}
 

internal/app/app.go

@@ -217,6 +217,22 @@ func (a *App) GetTreeSession() *session.TreeManager {
 	return a.opts.TreeSession
 }
 
+// SwitchTreeSession replaces the active tree session with a new one and
+// reloads the in-memory message store from the new session's messages.
+// The old tree session is closed. Used by /resume to switch sessions.
+func (a *App) SwitchTreeSession(ts *session.TreeManager) {
+	// Close old session.
+	if old := a.opts.TreeSession; old != nil {
+		_ = old.Close()
+	}
+	a.opts.TreeSession = ts
+	// Reload messages from new session.
+	a.store.Clear()
+	if ts != nil {
+		a.store.Replace(ts.GetFantasyMessages())
+	}
+}
+
 // AddContextMessage adds a user-role message to the conversation history
 // without triggering an LLM response. Used by the ! shell command prefix
 // to inject command output into context so the LLM can reference it in
@@ -391,41 +407,63 @@ func (a *App) Close() {
 // Internal: queue drain loop
 // --------------------------------------------------------------------------
 
-// drainQueue runs in a goroutine. It executes the given item and then
-// continues draining the queue until it is empty.
+// drainQueue runs in a goroutine. It collects all queued items (including the
+// first one) and submits them together as a single batch. This ensures that
+// when multiple messages are queued while the agent is working, they are all
+// submitted together in one turn rather than sequentially.
 // Must be called with a.busy == true and a.wg incremented.
 func (a *App) drainQueue(first queueItem) {
 	defer a.wg.Done()
 
-	item := first
-	for {
-		a.runQueueItem(item)
+	// Collect all items to process in this batch
+	var items []queueItem
+	items = append(items, first)
 
+	// Process batches until no more items are queued
+	for {
+		// Drain the queue to collect any pending items
 		a.mu.Lock()
-		// Stop draining if the app is shutting down.
-		if a.closed || a.rootCtx.Err() != nil {
-			a.busy = false
-			a.queue = a.queue[:0]
-			a.mu.Unlock()
-			return
-		}
-		if len(a.queue) == 0 {
-			a.busy = false
-			a.mu.Unlock()
-			return
-		}
-		item = a.queue[0]
-		a.queue = a.queue[1:]
-		qLen := len(a.queue)
+		items = append(items, a.queue...)
+		a.queue = a.queue[:0] // Clear the queue
+		queueLen := len(a.queue)
 		a.mu.Unlock()
-		// sendEvent must be called without a.mu held (see sendEvent comment).
-		a.sendEvent(QueueUpdatedEvent{Length: qLen})
+
+		// Send queue updated event (queue is now empty)
+		a.sendEvent(QueueUpdatedEvent{Length: queueLen})
+
+		// Process all collected items as a single batch
+		a.runQueueBatch(items)
+
+		// Check if more items were queued while we were processing
+		a.mu.Lock()
+		hasMore := len(a.queue) > 0
+		if hasMore {
+			// Start a new batch with the newly queued items
+			items = a.queue
+			a.queue = a.queue[:0]
+		}
+		a.mu.Unlock()
+
+		if !hasMore {
+			// No more items, we're done
+			break
+		}
+		// Process the new batch
 	}
+
+	// Mark as no longer busy
+	a.mu.Lock()
+	a.busy = false
+	a.mu.Unlock()
 }
 
-// runQueueItem executes a single queue item: adds the user message to the store,
-// runs the agent step, and sends the appropriate event to the program.
-func (a *App) runQueueItem(item queueItem) {
+// runQueueBatch executes multiple queue items as a single agent turn.
+// All items are submitted together, and the agent responds once to the combined context.
+func (a *App) runQueueBatch(items []queueItem) {
+	if len(items) == 0 {
+		return
+	}
+
 	// Create a per-step cancellable context.
 	stepCtx, cancel := context.WithCancel(a.rootCtx)
 	a.mu.Lock()
@@ -444,12 +482,17 @@ func (a *App) runQueueItem(item queueItem) {
 		}
 	}
 
-	result, err := a.executeStep(stepCtx, item.Prompt, eventFn, item.Files)
+	// Execute the batch
+	result, err := a.executeBatch(stepCtx, items, eventFn)
 	if err != nil {
 		if stepCtx.Err() != nil {
-			// Step was cancelled by the user (e.g. double-ESC). Send a
-			// cancellation event so the TUI can cut off the response
-			// cleanly without printing an error.
+			// Step was cancelled by the user (double-ESC). The SDK
+			// preserves the user message and any completed tool
+			// call/result pairs; only the in-progress message or tool
+			// call is discarded. Sync the in-memory store to match.
+			if ts := a.opts.TreeSession; ts != nil {
+				a.store.Replace(ts.GetFantasyMessages())
+			}
 			a.sendEvent(StepCancelledEvent{})
 			return
 		}
@@ -507,9 +550,87 @@ func (a *App) executeStep(ctx context.Context, prompt string, eventFn func(tea.M
 	return result, nil
 }
 
-// --------------------------------------------------------------------------
-// Internal: event helpers
-// --------------------------------------------------------------------------
+// executeBatch runs a batch of queue items as a single agent step by delegating
+// to the SDK's PromptResultWithMessages(), which handles session persistence,
+// hooks, extension events, and the generation loop.
+func (a *App) executeBatch(ctx context.Context, items []queueItem, eventFn func(tea.Msg)) (*kit.TurnResult, error) {
+	// Test hook: bypass SDK entirely (single item only for test compatibility).
+	if a.opts.PromptFunc != nil {
+		if len(items) == 1 {
+			return a.opts.PromptFunc(ctx, items[0].Prompt)
+		}
+		// For batch mode with PromptFunc, just use the first item
+		return a.opts.PromptFunc(ctx, items[0].Prompt)
+	}
+
+	sendFn := func(msg tea.Msg) {
+		if eventFn != nil {
+			eventFn(msg)
+		}
+	}
+
+	// Subscribe to SDK events for TUI rendering. The subscription is
+	// temporary — it lives only for the duration of this step.
+	unsub := a.subscribeSDKEvents(sendFn)
+	defer unsub()
+
+	// Show spinner while the agent works.
+	sendFn(SpinnerEvent{Show: true})
+
+	// Check if any items have file attachments
+	hasFiles := false
+	for _, item := range items {
+		if len(item.Files) > 0 {
+			hasFiles = true
+			break
+		}
+	}
+
+	var result *kit.TurnResult
+	var err error
+
+	if len(items) == 1 {
+		// Single item: use the original path for compatibility
+		item := items[0]
+		if len(item.Files) > 0 || hasFiles {
+			result, err = a.opts.Kit.PromptResultWithFiles(ctx, item.Prompt, item.Files)
+		} else {
+			result, err = a.opts.Kit.PromptResult(ctx, item.Prompt)
+		}
+	} else {
+		// Multiple items: batch them together
+		var messages []string
+		for _, item := range items {
+			messages = append(messages, item.Prompt)
+		}
+
+		// TODO: Handle file attachments in batch mode
+		// For now, files are ignored in batch mode (rare edge case)
+		if hasFiles {
+			// If files exist, fall back to processing just the first item with files
+			for _, item := range items {
+				if len(item.Files) > 0 {
+					result, err = a.opts.Kit.PromptResultWithFiles(ctx, item.Prompt, item.Files)
+					break
+				}
+			}
+		} else {
+			result, err = a.opts.Kit.PromptResultWithMessages(ctx, messages)
+		}
+	}
+
+	if err != nil {
+		return nil, err
+	}
+
+	// Sync in-memory store with the SDK's authoritative conversation.
+	a.store.Replace(result.Messages)
+
+	// Update usage tracker (using last item's prompt for tracking).
+	a.updateUsageFromTurnResult(result, items[len(items)-1].Prompt)
+
+	return result, nil
+}
 
 // sendEvent sends a tea.Msg to the registered program if one is set.
 // Must NOT be called with a.mu held (to avoid deadlock with the program).
@@ -550,6 +671,13 @@ func (a *App) subscribeSDKEvents(sendFn func(tea.Msg)) func() {
 			sendFn(StreamChunkEvent{Content: ev.Chunk})
 		case kit.ReasoningDeltaEvent:
 			sendFn(ReasoningChunkEvent{Delta: ev.Delta})
+		case kit.ToolOutputEvent:
+			sendFn(ToolOutputEvent{
+				ToolCallID: ev.ToolCallID,
+				ToolName:   ev.ToolName,
+				Chunk:      ev.Chunk,
+				IsStderr:   ev.IsStderr,
+			})
 		}
 	}))
 

internal/app/app_test.go

@@ -120,9 +120,8 @@ func TestRun_single(t *testing.T) {
 // Run (queued prompts)
 // --------------------------------------------------------------------------
 
-// TestRun_queued verifies that a second Run() call while the first is in-flight
-// enqueues the prompt rather than spawning a second goroutine, and that the
-// queue is drained after the first step completes.
+// TestRun_queued verifies that queued prompts are batched together and submitted
+// as a single agent turn rather than individually.
 func TestRun_queued(t *testing.T) {
 	gate := make(chan struct{})
 	callCount := 0
@@ -134,13 +133,7 @@ func TestRun_queued(t *testing.T) {
 			callCount++
 			mu.Unlock()
 			<-gate
-			return turnResult("first"), nil
-		},
-		func(_ context.Context) (*kit.TurnResult, error) {
-			mu.Lock()
-			callCount++
-			mu.Unlock()
-			return turnResult("second"), nil
+			return turnResult("batch result"), nil
 		},
 	)
 	app := newTestApp(stub)
@@ -165,11 +158,15 @@ func TestRun_queued(t *testing.T) {
 		t.Fatal("app did not become idle within 3s after queued runs")
 	}
 
+	// Wait for the goroutine to fully finish (avoid race with queue check)
+	app.wg.Wait()
+
 	mu.Lock()
 	total := callCount
 	mu.Unlock()
-	if total != 2 {
-		t.Fatalf("expected 2 calls, got %d", total)
+	// With batching, both prompts should be processed in a single call
+	if total != 1 {
+		t.Fatalf("expected 1 batched call, got %d", total)
 	}
 	if got := app.QueueLength(); got != 0 {
 		t.Fatalf("expected empty queue after drain, got %d", got)
@@ -180,31 +177,22 @@ func TestRun_queued(t *testing.T) {
 // Queue drain ordering
 // --------------------------------------------------------------------------
 
-// TestQueueDrainOrdering verifies that queued prompts are consumed in FIFO order.
+// TestQueueDrainOrdering verifies that queued prompts are batched together and
+// processed in a single agent turn.
 func TestQueueDrainOrdering(t *testing.T) {
 	gate := make(chan struct{})
-	var order []string
+	var receivedPrompt string
 	var mu sync.Mutex
 
 	stub := newStubWithFuncs(
 		func(ctx context.Context) (*kit.TurnResult, error) {
 			mu.Lock()
-			order = append(order, "first")
+			// In test mode with PromptFunc, we receive the first prompt
+			// but all messages are batched together
+			receivedPrompt = "batched"
 			mu.Unlock()
 			<-gate
-			return turnResult("first"), nil
-		},
-		func(_ context.Context) (*kit.TurnResult, error) {
-			mu.Lock()
-			order = append(order, "second")
-			mu.Unlock()
-			return turnResult("second"), nil
-		},
-		func(_ context.Context) (*kit.TurnResult, error) {
-			mu.Lock()
-			order = append(order, "third")
-			mu.Unlock()
-			return turnResult("third"), nil
+			return turnResult("batch result"), nil
 		},
 	)
 
@@ -228,16 +216,12 @@ func TestQueueDrainOrdering(t *testing.T) {
 	}
 
 	mu.Lock()
-	got := order
+	got := receivedPrompt
 	mu.Unlock()
 
-	if len(got) != 3 {
-		t.Fatalf("expected 3 calls, got %d: %v", len(got), got)
-	}
-	for i, want := range []string{"first", "second", "third"} {
-		if got[i] != want {
-			t.Fatalf("call[%d]: expected %q, got %q", i, want, got[i])
-		}
+	// With batching, all 3 prompts should be processed in a single call
+	if got != "batched" {
+		t.Fatalf("expected batched processing, got %q", got)
 	}
 }
 

internal/app/events.go

@@ -54,6 +54,19 @@ type ToolResultEvent struct {
 	IsError bool
 }
 
+// ToolOutputEvent is sent when a tool produces streaming output chunks (e.g., bash output).
+// This allows the TUI to display tool output as it arrives, before the tool completes.
+type ToolOutputEvent struct {
+	// ToolCallID is the stable identifier for the tool call producing output.
+	ToolCallID string
+	// ToolName is the name of the tool producing output.
+	ToolName string
+	// Chunk is a piece of the tool's output text.
+	Chunk string
+	// IsStderr indicates whether this chunk came from stderr.
+	IsStderr bool
+}
+
 // ToolCallContentEvent is sent when a step includes text content alongside tool calls.
 // This allows the TUI to display assistant commentary that accompanies tool usage.
 type ToolCallContentEvent struct {

internal/compaction/compaction.go

@@ -5,10 +5,18 @@
 // messages (KeepRecentTokens, default 20 000) rather than a fixed message
 // count. Auto-compaction fires when estimated context usage exceeds
 // contextWindow − ReserveTokens.
+//
+// Features modelled after pi's compaction system:
+//   - Tool result truncation (2000 char max) during serialisation
+//   - Split turn handling: when a single turn exceeds the keep budget,
+//     the turn prefix is summarised separately and merged
+//   - Cumulative file tracking: read and modified files extracted from
+//     tool calls and carried forward across compactions
 package compaction
 
 import (
 	"context"
+	"encoding/json"
 	"fmt"
 	"strings"
 
@@ -66,10 +74,13 @@ func ShouldCompact(messages []fantasy.Message, contextWindow int, reserveTokens
 
 // CompactionResult contains statistics from a compaction operation.
 type CompactionResult struct {
-	Summary         string // LLM-generated summary of compacted messages
-	OriginalTokens  int    // Estimated token count before compaction
-	CompactedTokens int    // Estimated token count after compaction
-	MessagesRemoved int    // Number of messages replaced by the summary
+	Summary         string   // LLM-generated summary of compacted messages
+	OriginalTokens  int      // Estimated token count before compaction
+	CompactedTokens int      // Estimated token count after compaction
+	MessagesRemoved int      // Number of messages replaced by the summary
+	CutPoint        int      // Index in the original messages where the cut was made
+	ReadFiles       []string // Files read during the compacted conversation
+	ModifiedFiles   []string // Files modified during the compacted conversation
 }
 
 // CompactionOptions configures compaction behaviour. Token-based defaults
@@ -130,8 +141,34 @@ Use this EXACT format:
 - [Any data, examples, or references needed to continue]
 - [Or "(none)" if not applicable]
 
+<read-files>
+[One file path per line for files that were read during the conversation]
+</read-files>
+
+<modified-files>
+[One file path per line for files that were created, edited, or written during the conversation]
+</modified-files>
+
 Keep each section concise. Preserve exact file paths, function names, and error messages.`
 
+// ---------------------------------------------------------------------------
+// Tool result truncation
+// ---------------------------------------------------------------------------
+
+// maxToolResultChars is the maximum length of tool result text preserved
+// during serialisation. Longer results are truncated with a marker.
+const maxToolResultChars = 2000
+
+// truncateToolResult truncates text to maxToolResultChars, appending a
+// marker indicating how many characters were removed.
+func truncateToolResult(text string) string {
+	if len(text) <= maxToolResultChars {
+		return text
+	}
+	truncated := len(text) - maxToolResultChars
+	return text[:maxToolResultChars] + fmt.Sprintf("\n[...%d chars truncated]", truncated)
+}
+
 // ---------------------------------------------------------------------------
 // Cut point (token-based)
 // ---------------------------------------------------------------------------
@@ -143,11 +180,26 @@ func isValidCutPoint(msg fantasy.Message) bool {
 	return msg.Role != fantasy.MessageRoleTool
 }
 
+// findTurnStart returns the index of the user message that starts the turn
+// containing messages[idx]. A "turn" starts with a user message and includes
+// all subsequent assistant/tool messages until the next user message.
+func findTurnStart(messages []fantasy.Message, idx int) int {
+	for i := idx; i >= 0; i-- {
+		if messages[i].Role == fantasy.MessageRoleUser {
+			return i
+		}
+	}
+	return 0
+}
+
 // FindCutPoint walks backward from the end of messages, accumulating tokens
 // until the keepRecentTokens budget is filled. Returns the index that
 // separates "old" messages (0..cutPoint-1, to be summarised) from "recent"
 // messages (cutPoint..end, to be preserved).
 //
+// The cut point prefers turn boundaries (user messages). When a single turn
+// exceeds the budget, the cut lands mid-turn (IsSplitTurn returns true).
+//
 // Returns 0 if there are fewer than 2 messages or all messages fit within
 // the keep budget.
 func FindCutPoint(messages []fantasy.Message, keepRecentTokens int) int {
@@ -193,6 +245,23 @@ func FindCutPoint(messages []fantasy.Message, keepRecentTokens int) int {
 	return 0
 }
 
+// IsSplitTurn returns true if the cut point lands in the middle of a turn
+// (i.e. the message at cutPoint is not a user message, meaning we're
+// splitting a single turn's assistant/tool messages).
+func IsSplitTurn(messages []fantasy.Message, cutPoint int) bool {
+	if cutPoint <= 0 || cutPoint >= len(messages) {
+		return false
+	}
+	// If the cut point is at a user message, it's a clean turn boundary.
+	if messages[cutPoint].Role == fantasy.MessageRoleUser {
+		return false
+	}
+	// Otherwise we're cutting mid-turn — check if the turn started before
+	// the cut point.
+	turnStart := findTurnStart(messages, cutPoint)
+	return turnStart < cutPoint
+}
+
 // forceCutPoint returns a cut point that keeps only the last non-tool
 // message, summarising everything before it. Used when the budget-based
 // FindCutPoint returns 0 but the caller wants to compact anyway (manual
@@ -207,12 +276,104 @@ func forceCutPoint(messages []fantasy.Message) int {
 	return 0
 }
 
+// ---------------------------------------------------------------------------
+// File tracking
+// ---------------------------------------------------------------------------
+
+// fileOps contains cumulative file operation tracking.
+type fileOps struct {
+	ReadFiles     map[string]bool
+	ModifiedFiles map[string]bool
+}
+
+func newFileOps() *fileOps {
+	return &fileOps{
+		ReadFiles:     make(map[string]bool),
+		ModifiedFiles: make(map[string]bool),
+	}
+}
+
+// extractFileOps scans messages for tool calls and extracts file paths.
+// It recognises the built-in Kit tools: read, write, edit, bash, grep, find, ls.
+func extractFileOps(messages []fantasy.Message) *fileOps {
+	ops := newFileOps()
+	for _, msg := range messages {
+		for _, part := range msg.Content {
+			tc, ok := part.(fantasy.ToolCallPart)
+			if !ok {
+				continue
+			}
+
+			// Parse the JSON input to extract path arguments.
+			var args map[string]any
+			if err := json.Unmarshal([]byte(tc.Input), &args); err != nil {
+				continue
+			}
+
+			path, _ := args["path"].(string)
+			if path == "" {
+				continue
+			}
+
+			switch tc.ToolName {
+			case "read", "grep", "find", "ls":
+				ops.ReadFiles[path] = true
+			case "write", "edit":
+				ops.ModifiedFiles[path] = true
+			}
+		}
+	}
+	return ops
+}
+
+// merge combines another fileOps into this one (for cumulative tracking).
+func (f *fileOps) merge(other *fileOps) {
+	if other == nil {
+		return
+	}
+	for k := range other.ReadFiles {
+		f.ReadFiles[k] = true
+	}
+	for k := range other.ModifiedFiles {
+		f.ModifiedFiles[k] = true
+	}
+}
+
+// mergeSlices adds previously tracked file lists (from a prior compaction).
+func (f *fileOps) mergeSlices(readFiles, modifiedFiles []string) {
+	for _, p := range readFiles {
+		f.ReadFiles[p] = true
+	}
+	for _, p := range modifiedFiles {
+		f.ModifiedFiles[p] = true
+	}
+}
+
+// sortedKeys returns the keys of a bool map sorted alphabetically.
+func sortedKeys(m map[string]bool) []string {
+	if len(m) == 0 {
+		return nil
+	}
+	keys := make([]string, 0, len(m))
+	for k := range m {
+		keys = append(keys, k)
+	}
+	// Simple sort — no need for sort package for small lists.
+	for i := 0; i < len(keys); i++ {
+		for j := i + 1; j < len(keys); j++ {
+			if keys[j] < keys[i] {
+				keys[i], keys[j] = keys[j], keys[i]
+			}
+		}
+	}
+	return keys
+}
+
 // ---------------------------------------------------------------------------
 // Message serialisation
 // ---------------------------------------------------------------------------
 
-// roleLabel returns a human-readable label for a fantasy message role,
-
+// roleLabel returns a human-readable label for a fantasy message role.
 func roleLabel(role fantasy.MessageRole) string {
 	switch role {
 	case fantasy.MessageRoleUser:
@@ -229,16 +390,26 @@ func roleLabel(role fantasy.MessageRole) string {
 }
 
 // serializeMessages converts a slice of fantasy messages into a plain-text
-// representation suitable for sending to the summarisation LLM. The format
-
+// representation suitable for sending to the summarisation LLM. Tool result
+// text is truncated to maxToolResultChars to keep the summarisation request
+// within reasonable token budgets.
 func serializeMessages(messages []fantasy.Message) string {
 	var sb strings.Builder
 	for _, msg := range messages {
 		sb.WriteString(roleLabel(msg.Role))
 		sb.WriteString(":\n")
 		for _, part := range msg.Content {
-			if tp, ok := part.(fantasy.TextPart); ok {
-				sb.WriteString(tp.Text)
+			switch p := part.(type) {
+			case fantasy.TextPart:
+				if msg.Role == fantasy.MessageRoleTool {
+					sb.WriteString(truncateToolResult(p.Text))
+				} else {
+					sb.WriteString(p.Text)
+				}
+			case fantasy.ToolCallPart:
+				fmt.Fprintf(&sb, "[Tool call: %s(%s)]", p.ToolName, truncateToolResult(p.Input))
+			case fantasy.ReasoningPart:
+				fmt.Fprintf(&sb, "[Thinking]: %s", truncateToolResult(p.Text))
 			}
 		}
 		sb.WriteString("\n\n")
@@ -250,6 +421,13 @@ func serializeMessages(messages []fantasy.Message) string {
 // Compact
 // ---------------------------------------------------------------------------
 
+// PreviousCompaction carries file tracking state from a prior compaction so
+// that file operations accumulate across multiple compactions.
+type PreviousCompaction struct {
+	ReadFiles     []string
+	ModifiedFiles []string
+}
+
 // Compact summarises older messages using the LLM, returning the compaction
 // result and a new message slice (summary message + preserved recent
 // messages).
@@ -261,12 +439,16 @@ func serializeMessages(messages []fantasy.Message) string {
 // customInstructions is optional text appended to the summary prompt (e.g.
 // "Focus on the API design decisions"). Pass "" to use the default prompt
 // only.
+//
+// prev carries file tracking from a previous compaction for cumulative
+// tracking. Pass nil if there is no prior compaction.
 func Compact(
 	ctx context.Context,
 	model fantasy.LanguageModel,
 	messages []fantasy.Message,
 	opts CompactionOptions,
 	customInstructions string,
+	prev *PreviousCompaction,
 ) (*CompactionResult, []fantasy.Message, error) {
 	opts.defaults()
 
@@ -289,30 +471,30 @@ func Compact(
 	recentMessages := messages[cutPoint:]
 	originalTokens := EstimateMessageTokens(messages)
 
-	// Serialise old messages to text.
-	conversationText := serializeMessages(oldMessages)
-
-	// Build the user-facing prompt: conversation text + summary instructions.
-	userPrompt := opts.SummaryPrompt
-	if userPrompt == "" {
-		userPrompt = defaultSummaryPrompt
-	}
-	if customInstructions != "" {
-		userPrompt += "\n\nAdditional instructions: " + customInstructions
+	// Extract file operations from old messages.
+	ops := extractFileOps(oldMessages)
+	// Accumulate from previous compaction if present.
+	if prev != nil {
+		ops.mergeSlices(prev.ReadFiles, prev.ModifiedFiles)
 	}
+	// Also scan recent messages for file ops (they'll be carried forward).
+	recentOps := extractFileOps(recentMessages)
+	ops.merge(recentOps)
 
-	// Create a lightweight agent (no tools) just for summarisation.
-	summaryAgent := fantasy.NewAgent(model,
-		fantasy.WithSystemPrompt(defaultSystemPrompt),
-	)
-	result, err := summaryAgent.Generate(ctx, fantasy.AgentCall{
-		Prompt: conversationText + "\n\n" + userPrompt,
-	})
+	// Handle split turns: when the cut lands mid-turn, summarise the turn
+	// prefix separately and merge with the history summary.
+	var summaryText string
+	var err error
+
+	if IsSplitTurn(messages, cutPoint) {
+		summaryText, err = compactSplitTurn(ctx, model, oldMessages, messages, cutPoint, opts, customInstructions)
+	} else {
+		summaryText, err = compactNormal(ctx, model, oldMessages, opts, customInstructions)
+	}
 	if err != nil {
-		return nil, nil, fmt.Errorf("compaction summarisation failed: %w", err)
+		return nil, nil, err
 	}
 
-	summaryText := result.Response.Content.Text()
 	if summaryText == "" {
 		return nil, nil, fmt.Errorf("compaction produced an empty summary")
 	}
@@ -338,5 +520,120 @@ func Compact(
 		OriginalTokens:  originalTokens,
 		CompactedTokens: compactedTokens,
 		MessagesRemoved: len(oldMessages),
+		CutPoint:        cutPoint,
+		ReadFiles:       sortedKeys(ops.ReadFiles),
+		ModifiedFiles:   sortedKeys(ops.ModifiedFiles),
 	}, newMessages, nil
 }
+
+// compactNormal generates a summary for a clean turn-boundary cut.
+func compactNormal(
+	ctx context.Context,
+	model fantasy.LanguageModel,
+	oldMessages []fantasy.Message,
+	opts CompactionOptions,
+	customInstructions string,
+) (string, error) {
+	conversationText := serializeMessages(oldMessages)
+	return generateSummary(ctx, model, conversationText, opts, customInstructions)
+}
+
+// compactSplitTurn handles the case where the cut point lands mid-turn.
+// It generates two summaries and merges them:
+//  1. History summary: all complete turns before the split turn
+//  2. Turn prefix summary: the early part of the split turn (from the turn's
+//     user message up to the cut point)
+//
+// The merged result preserves context from both the older history and the
+// beginning of the current long turn.
+func compactSplitTurn(
+	ctx context.Context,
+	model fantasy.LanguageModel,
+	oldMessages []fantasy.Message,
+	allMessages []fantasy.Message,
+	cutPoint int,
+	opts CompactionOptions,
+	customInstructions string,
+) (string, error) {
+	// Find where the split turn starts.
+	turnStart := findTurnStart(allMessages, cutPoint)
+
+	// Messages before the turn are the "history" portion.
+	historyMessages := oldMessages
+	if turnStart > 0 && turnStart < len(oldMessages) {
+		historyMessages = oldMessages[:turnStart]
+	}
+
+	// The turn prefix: from turnStart to cutPoint.
+	turnPrefixMessages := allMessages[turnStart:cutPoint]
+
+	var historySummary string
+	var err error
+
+	// Generate history summary if there are complete turns before the split.
+	if len(historyMessages) >= 2 {
+		historySummary, err = generateSummary(ctx, model,
+			serializeMessages(historyMessages), opts, "")
+		if err != nil {
+			return "", fmt.Errorf("split turn history summary failed: %w", err)
+		}
+	}
+
+	// Generate turn prefix summary.
+	turnPrefixText := serializeMessages(turnPrefixMessages)
+	turnPrefixPrompt := "The messages above are the BEGINNING of a long turn that was split. " +
+		"Summarize the work done so far in this turn, preserving tool call results, " +
+		"file changes, and progress. Another LLM will continue this turn."
+	if customInstructions != "" {
+		turnPrefixPrompt += "\n\nAdditional instructions: " + customInstructions
+	}
+
+	summaryAgent := fantasy.NewAgent(model,
+		fantasy.WithSystemPrompt(defaultSystemPrompt),
+	)
+	result, err := summaryAgent.Generate(ctx, fantasy.AgentCall{
+		Prompt: turnPrefixText + "\n\n" + turnPrefixPrompt,
+	})
+	if err != nil {
+		return "", fmt.Errorf("split turn prefix summary failed: %w", err)
+	}
+	turnPrefixSummary := result.Response.Content.Text()
+
+	// Merge the two summaries.
+	if historySummary != "" && turnPrefixSummary != "" {
+		return historySummary + "\n\n---\n\n## Current Turn (in progress)\n\n" + turnPrefixSummary, nil
+	}
+	if turnPrefixSummary != "" {
+		return turnPrefixSummary, nil
+	}
+	return historySummary, nil
+}
+
+// generateSummary calls the LLM to produce a structured summary.
+func generateSummary(
+	ctx context.Context,
+	model fantasy.LanguageModel,
+	conversationText string,
+	opts CompactionOptions,
+	customInstructions string,
+) (string, error) {
+	userPrompt := opts.SummaryPrompt
+	if userPrompt == "" {
+		userPrompt = defaultSummaryPrompt
+	}
+	if customInstructions != "" {
+		userPrompt += "\n\nAdditional instructions: " + customInstructions
+	}
+
+	summaryAgent := fantasy.NewAgent(model,
+		fantasy.WithSystemPrompt(defaultSystemPrompt),
+	)
+	result, err := summaryAgent.Generate(ctx, fantasy.AgentCall{
+		Prompt: conversationText + "\n\n" + userPrompt,
+	})
+	if err != nil {
+		return "", fmt.Errorf("compaction summarisation failed: %w", err)
+	}
+
+	return result.Response.Content.Text(), nil
+}

internal/compaction/compaction_test.go

@@ -243,7 +243,7 @@ func TestCompact_TooFewMessages(t *testing.T) {
 		makeTextMessageN(fantasy.MessageRoleUser, 400),
 	}
 
-	result, newMsgs, err := Compact(context.TODO(), nil, msgs, CompactionOptions{}, "")
+	result, newMsgs, err := Compact(context.TODO(), nil, msgs, CompactionOptions{}, "", nil)
 	if err != nil {
 		t.Fatalf("unexpected error: %v", err)
 	}
@@ -262,7 +262,7 @@ func TestCompact_WithinBudget(t *testing.T) {
 		makeTextMessageN(fantasy.MessageRoleAssistant, 400),
 	}
 
-	result, newMsgs, err := Compact(context.TODO(), nil, msgs, CompactionOptions{}, "")
+	result, newMsgs, err := Compact(context.TODO(), nil, msgs, CompactionOptions{}, "", nil)
 	if err != nil {
 		t.Fatalf("unexpected error: %v", err)
 	}
@@ -273,3 +273,169 @@ func TestCompact_WithinBudget(t *testing.T) {
 		t.Errorf("messages changed: got %d, want %d", len(newMsgs), len(msgs))
 	}
 }
+
+// ---------------------------------------------------------------------------
+// Tool result truncation
+// ---------------------------------------------------------------------------
+
+func TestTruncateToolResult(t *testing.T) {
+	// Short text — no truncation.
+	short := strings.Repeat("x", 100)
+	if got := truncateToolResult(short); got != short {
+		t.Errorf("truncated short text unexpectedly")
+	}
+
+	// Exactly at limit.
+	exact := strings.Repeat("x", maxToolResultChars)
+	if got := truncateToolResult(exact); got != exact {
+		t.Errorf("truncated text at exact limit")
+	}
+
+	// Over limit.
+	over := strings.Repeat("x", maxToolResultChars+500)
+	got := truncateToolResult(over)
+	if len(got) > maxToolResultChars+50 { // allow room for marker
+		t.Errorf("truncated text too long: %d chars", len(got))
+	}
+	if !strings.Contains(got, "500 chars truncated") {
+		t.Errorf("truncation marker missing, got: %s", got[maxToolResultChars:])
+	}
+}
+
+func TestSerializeMessages_TruncatesToolResults(t *testing.T) {
+	longResult := strings.Repeat("R", maxToolResultChars+1000)
+	msgs := []fantasy.Message{
+		makeTextMessage(fantasy.MessageRoleUser, "question"),
+		{
+			Role:    fantasy.MessageRoleTool,
+			Content: []fantasy.MessagePart{fantasy.TextPart{Text: longResult}},
+		},
+	}
+
+	serialized := serializeMessages(msgs)
+	if strings.Contains(serialized, longResult) {
+		t.Error("tool result was not truncated during serialisation")
+	}
+	if !strings.Contains(serialized, "chars truncated") {
+		t.Error("truncation marker missing in serialised output")
+	}
+}
+
+func TestSerializeMessages_PreservesNonToolText(t *testing.T) {
+	longText := strings.Repeat("T", maxToolResultChars+1000)
+	msgs := []fantasy.Message{
+		makeTextMessage(fantasy.MessageRoleUser, longText),
+	}
+
+	serialized := serializeMessages(msgs)
+	if !strings.Contains(serialized, longText) {
+		t.Error("non-tool text was unexpectedly truncated")
+	}
+}
+
+// ---------------------------------------------------------------------------
+// Split turn detection
+// ---------------------------------------------------------------------------
+
+func TestIsSplitTurn(t *testing.T) {
+	msgs := []fantasy.Message{
+		makeTextMessageN(fantasy.MessageRoleUser, 400),      // 0: turn 1 user
+		makeTextMessageN(fantasy.MessageRoleAssistant, 400), // 1: turn 1 assistant
+		makeTextMessageN(fantasy.MessageRoleUser, 400),      // 2: turn 2 user
+		makeTextMessageN(fantasy.MessageRoleAssistant, 400), // 3: turn 2 assistant
+		makeTextMessageN(fantasy.MessageRoleTool, 400),      // 4: turn 2 tool result
+		makeTextMessageN(fantasy.MessageRoleAssistant, 400), // 5: turn 2 assistant
+	}
+
+	tests := []struct {
+		name     string
+		cutPoint int
+		want     bool
+	}{
+		{"at user message (turn boundary)", 2, false},
+		{"at assistant mid-turn", 3, true},
+		{"at assistant after tool (mid-turn)", 5, true},
+		{"at 0 (no cut)", 0, false},
+		{"beyond range", 10, false},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			got := IsSplitTurn(msgs, tt.cutPoint)
+			if got != tt.want {
+				t.Errorf("IsSplitTurn(msgs, %d) = %v, want %v", tt.cutPoint, got, tt.want)
+			}
+		})
+	}
+}
+
+// ---------------------------------------------------------------------------
+// File operations extraction
+// ---------------------------------------------------------------------------
+
+func TestExtractFileOps(t *testing.T) {
+	// Create messages with tool calls.
+	msgs := []fantasy.Message{
+		{
+			Role: fantasy.MessageRoleAssistant,
+			Content: []fantasy.MessagePart{
+				fantasy.ToolCallPart{ToolCallID: "1", ToolName: "read", Input: `{"path":"src/main.go"}`},
+				fantasy.ToolCallPart{ToolCallID: "2", ToolName: "write", Input: `{"path":"src/out.go"}`},
+				fantasy.ToolCallPart{ToolCallID: "3", ToolName: "edit", Input: `{"path":"src/edit.go"}`},
+				fantasy.ToolCallPart{ToolCallID: "4", ToolName: "grep", Input: `{"path":"src/search"}`},
+			},
+		},
+	}
+
+	ops := extractFileOps(msgs)
+	if !ops.ReadFiles["src/main.go"] {
+		t.Error("read file not tracked: src/main.go")
+	}
+	if !ops.ReadFiles["src/search"] {
+		t.Error("grep path not tracked as read: src/search")
+	}
+	if !ops.ModifiedFiles["src/out.go"] {
+		t.Error("write file not tracked: src/out.go")
+	}
+	if !ops.ModifiedFiles["src/edit.go"] {
+		t.Error("edit file not tracked: src/edit.go")
+	}
+}
+
+func TestFileOps_MergeSlices(t *testing.T) {
+	ops := newFileOps()
+	ops.ReadFiles["a.go"] = true
+	ops.ModifiedFiles["b.go"] = true
+
+	ops.mergeSlices(
+		[]string{"c.go", "a.go"},
+		[]string{"d.go"},
+	)
+
+	if len(ops.ReadFiles) != 2 { // a.go, c.go
+		t.Errorf("ReadFiles len = %d, want 2", len(ops.ReadFiles))
+	}
+	if len(ops.ModifiedFiles) != 2 { // b.go, d.go
+		t.Errorf("ModifiedFiles len = %d, want 2", len(ops.ModifiedFiles))
+	}
+}
+
+func TestSortedKeys(t *testing.T) {
+	m := map[string]bool{"c": true, "a": true, "b": true}
+	got := sortedKeys(m)
+	want := []string{"a", "b", "c"}
+	if len(got) != len(want) {
+		t.Fatalf("sortedKeys len = %d, want %d", len(got), len(want))
+	}
+	for i, v := range got {
+		if v != want[i] {
+			t.Errorf("sortedKeys[%d] = %q, want %q", i, v, want[i])
+		}
+	}
+}
+
+func TestSortedKeys_Empty(t *testing.T) {
+	got := sortedKeys(nil)
+	if got != nil {
+		t.Errorf("sortedKeys(nil) = %v, want nil", got)
+	}
+}

internal/config/config.go

@@ -183,6 +183,10 @@ type Config struct {
 
 	// TLS configuration
 	TLSSkipVerify bool `json:"tls-skip-verify,omitempty" yaml:"tls-skip-verify,omitempty"`
+
+	// Prompt templates configuration
+	Prompts           []string `json:"prompts,omitempty" yaml:"prompts,omitempty"`
+	NoPromptTemplates bool     `json:"no-prompt-templates,omitempty" yaml:"no-prompt-templates,omitempty"`
 }
 
 // GetTransportType returns the transport type for the server config, mapping

internal/core/bash.go

@@ -1,16 +1,41 @@
 package core
 
 import (
-	"bytes"
+	"bufio"
 	"context"
 	"fmt"
+	"io"
+	"os"
 	"os/exec"
 	"strings"
+	"sync"
 	"time"
 
 	"charm.land/fantasy"
 )
 
+// ToolOutputCallback is the signature for streaming tool output.
+// It receives tool call ID, tool name, output chunk, and whether it's stderr.
+type ToolOutputCallback func(toolCallID, toolName, chunk string, isStderr bool)
+
+// contextKey is a custom type for context keys to avoid collisions.
+type contextKey string
+
+const toolOutputCallbackKey contextKey = "toolOutputCallback"
+
+// ContextWithToolOutputCallback returns a new context with the tool output callback set.
+func ContextWithToolOutputCallback(ctx context.Context, callback ToolOutputCallback) context.Context {
+	return context.WithValue(ctx, toolOutputCallbackKey, callback)
+}
+
+// toolOutputCallbackFromContext retrieves the tool output callback from context.
+func toolOutputCallbackFromContext(ctx context.Context) ToolOutputCallback {
+	if cb, ok := ctx.Value(toolOutputCallbackKey).(ToolOutputCallback); ok {
+		return cb
+	}
+	return nil
+}
+
 const defaultBashTimeout = 120 * time.Second
 const maxBashTimeout = 600 * time.Second
 
@@ -90,7 +115,29 @@ func executeBash(ctx context.Context, call fantasy.ToolCall, workDir string) (fa
 		cmd.Dir = workDir
 	}
 
-	var stdout, stderr bytes.Buffer
+	// Ensure SHELL is set to bash so child processes (e.g. tmux) use bash
+	// rather than the user's login shell (which may be nushell, fish, etc.).
+	bashPath, err := exec.LookPath("bash")
+	if err != nil {
+		bashPath = "/bin/bash"
+	}
+	cmd.Env = append(os.Environ(), "SHELL="+bashPath)
+
+	// Get the output callback if present (for streaming support)
+	outputCallback := toolOutputCallbackFromContext(ctx)
+
+	if outputCallback != nil {
+		// Streaming mode: use pipes to capture output as it arrives
+		return executeBashStreaming(cmdCtx, call, cmd, outputCallback)
+	}
+
+	// Non-streaming mode: collect all output at once (original behavior)
+	return executeBashBuffered(cmdCtx, call, cmd)
+}
+
+// executeBashBuffered collects all output before returning (original behavior).
+func executeBashBuffered(cmdCtx context.Context, call fantasy.ToolCall, cmd *exec.Cmd) (fantasy.ToolResponse, error) {
+	var stdout, stderr strings.Builder
 	cmd.Stdout = &stdout
 	cmd.Stderr = &stderr
 
@@ -101,21 +148,93 @@ func executeBash(ctx context.Context, call fantasy.ToolCall, workDir string) (fa
 		if exitErr, ok := err.(*exec.ExitError); ok {
 			exitCode = exitErr.ExitCode()
 		} else if cmdCtx.Err() == context.DeadlineExceeded {
-			return fantasy.NewTextErrorResponse(fmt.Sprintf("command timed out after %v", timeout)), nil
+			return fantasy.NewTextErrorResponse("command timed out"), nil
 		}
 	}
 
-	// Build result
-	var result strings.Builder
-	if stdout.Len() > 0 {
-		result.WriteString(stdout.String())
+	return buildBashResponse(stdout.String(), stderr.String(), exitCode)
+}
+
+// executeBashStreaming streams output as it arrives via the callback.
+func executeBashStreaming(cmdCtx context.Context, call fantasy.ToolCall, cmd *exec.Cmd, outputCallback ToolOutputCallback) (fantasy.ToolResponse, error) {
+	stdoutPipe, err := cmd.StdoutPipe()
+	if err != nil {
+		return fantasy.NewTextErrorResponse("failed to create stdout pipe"), nil
 	}
-	if stderr.Len() > 0 {
+	stderrPipe, err := cmd.StderrPipe()
+	if err != nil {
+		return fantasy.NewTextErrorResponse("failed to create stderr pipe"), nil
+	}
+
+	// Start command execution
+	if err := cmd.Start(); err != nil {
+		return fantasy.NewTextErrorResponse(fmt.Sprintf("failed to start command: %v", err)), nil
+	}
+
+	// Stream stdout and stderr concurrently
+	var wg sync.WaitGroup
+	var stdoutChunks, stderrChunks []string
+
+	streamOutput := func(reader io.Reader, isStderr bool) {
+		defer wg.Done()
+		scanner := bufio.NewScanner(reader)
+		// Use larger buffer for long lines
+		buf := make([]byte, 0, 64*1024)
+		scanner.Buffer(buf, 1024*1024)
+
+		for scanner.Scan() {
+			chunk := scanner.Text()
+			// Send chunk to UI
+			outputCallback(call.ID, "bash", chunk, isStderr)
+			// Collect for final result
+			if isStderr {
+				stderrChunks = append(stderrChunks, chunk)
+			} else {
+				stdoutChunks = append(stdoutChunks, chunk)
+			}
+			// Check if context was cancelled
+			select {
+			case <-cmdCtx.Done():
+				return
+			default:
+			}
+		}
+	}
+
+	wg.Add(2)
+	go streamOutput(stdoutPipe, false)
+	go streamOutput(stderrPipe, true)
+
+	// Wait for both streams to complete
+	wg.Wait()
+
+	// Wait for command to finish
+	err = cmd.Wait()
+
+	exitCode := 0
+	if err != nil {
+		if exitErr, ok := err.(*exec.ExitError); ok {
+			exitCode = exitErr.ExitCode()
+		} else if cmdCtx.Err() == context.DeadlineExceeded {
+			return fantasy.NewTextErrorResponse("command timed out"), nil
+		}
+	}
+
+	return buildBashResponse(strings.Join(stdoutChunks, "\n"), strings.Join(stderrChunks, "\n"), exitCode)
+}
+
+// buildBashResponse constructs the final tool response from stdout/stderr.
+func buildBashResponse(stdout, stderr string, exitCode int) (fantasy.ToolResponse, error) {
+	var result strings.Builder
+	if stdout != "" {
+		result.WriteString(stdout)
+	}
+	if stderr != "" {
 		if result.Len() > 0 {
 			result.WriteString("\n")
 		}
 		result.WriteString("STDERR:\n")
-		result.WriteString(stderr.String())
+		result.WriteString(stderr)
 	}
 	if exitCode != 0 {
 		if result.Len() > 0 {

internal/core/edit.go

@@ -6,8 +6,11 @@ import (
 	"os"
 	"strings"
 	"unicode"
+	"unicode/utf8"
 
 	"charm.land/fantasy"
+
+	udiff "github.com/aymanbagabas/go-udiff"
 )
 
 type editArgs struct {
@@ -82,7 +85,7 @@ func executeEdit(ctx context.Context, call fantasy.ToolCall, workDir string) (fa
 			if err := os.WriteFile(absPath, []byte(newContent), 0644); err != nil {
 				return fantasy.NewTextErrorResponse(fmt.Sprintf("failed to write file: %v", err)), nil
 			}
-			diff := generateDiff(absPath, normalized, newContent, idx)
+			diff := generateDiff(absPath, normalized, newContent)
 			resp := fantasy.NewTextResponse(fmt.Sprintf("Applied edit (fuzzy match) to %s\n%s", args.Path, diff))
 			return fantasy.WithResponseMetadata(resp, editDiffMeta(absPath, matchedText, args.NewText)), nil
 		}
@@ -100,8 +103,7 @@ func executeEdit(ctx context.Context, call fantasy.ToolCall, workDir string) (fa
 		return fantasy.NewTextErrorResponse(fmt.Sprintf("failed to write file: %v", err)), nil
 	}
 
-	idx := strings.Index(normalized, normalizedOld)
-	diff := generateDiff(absPath, normalized, newContent, idx)
+	diff := generateDiff(absPath, normalized, newContent)
 	resp := fantasy.NewTextResponse(fmt.Sprintf("Applied edit to %s\n%s", args.Path, diff))
 	return fantasy.WithResponseMetadata(resp, editDiffMeta(absPath, normalizedOld, args.NewText)), nil
 }
@@ -122,102 +124,108 @@ func editDiffMeta(path, oldText, newText string) map[string]any {
 }
 
 // fuzzyMatch tries to find old_text with relaxed matching:
-// - Strips trailing whitespace per line
-// - Normalizes unicode quotes to ASCII
-// - Normalizes unicode dashes/spaces
-// Returns (index, matchLength) or (-1, 0) if not found.
+//   - Strips trailing whitespace per line
+//   - Normalizes unicode quotes to ASCII
+//   - Normalizes unicode dashes/spaces
+//
+// Returns (index, matchLength) in the original content, or (-1, 0) if not
+// found or ambiguous (multiple matches).
 func fuzzyMatch(content, search string) (int, int) {
-	normalizedContent := normalizeForFuzzy(content)
-	normalizedSearch := normalizeForFuzzy(search)
+	normContent, contentMap := normalizeWithMap(content)
+	normSearch := normalizeForFuzzy(search)
 
-	idx := strings.Index(normalizedContent, normalizedSearch)
+	if normSearch == "" {
+		return -1, 0
+	}
+
+	idx := strings.Index(normContent, normSearch)
 	if idx < 0 {
 		return -1, 0
 	}
 
-	// Map back to original content position
-	// Since normalization can change lengths, we need to find the
-	// corresponding region in the original content
-	origIdx := mapFuzzyIndex(content, normalizedContent, idx)
-	origEnd := mapFuzzyIndex(content, normalizedContent, idx+len(normalizedSearch))
+	// Reject ambiguous matches — if there are multiple fuzzy matches
+	// we can't safely pick one.
+	if strings.Count(normContent, normSearch) > 1 {
+		return -1, 0
+	}
 
-	return origIdx, origEnd - origIdx
+	// Map normalized byte positions back to original byte positions.
+	origStart := contentMap[idx]
+	endNorm := idx + len(normSearch)
+	var origEnd int
+	if endNorm >= len(normContent) {
+		origEnd = len(content)
+	} else {
+		origEnd = contentMap[endNorm]
+	}
+
+	return origStart, origEnd - origStart
 }
 
-func normalizeForFuzzy(s string) string {
-	// Strip trailing whitespace per line
+// normalizeWithMap normalizes s for fuzzy matching and returns both the
+// normalized string and a byte-position mapping where mapping[i] is the
+// original byte position corresponding to normalized byte position i.
+//
+// Normalization: trim trailing whitespace per line, replace unicode
+// quotes/dashes/spaces with their ASCII equivalents.
+func normalizeWithMap(s string) (string, []int) {
+	var result []byte
+	var mapping []int // mapping[i] = original byte position for result byte i
+
 	lines := strings.Split(s, "\n")
-	for i, line := range lines {
-		lines[i] = strings.TrimRightFunc(line, unicode.IsSpace)
-	}
-	result := strings.Join(lines, "\n")
-
-	// Normalize smart quotes
-	replacer := strings.NewReplacer(
-		"\u201c", "\"", // left double quote
-		"\u201d", "\"", // right double quote
-		"\u2018", "'", // left single quote
-		"\u2019", "'", // right single quote
-		"\u2013", "-", // en dash
-		"\u2014", "-", // em dash
-		"\u00a0", " ", // non-breaking space
-	)
-	return replacer.Replace(result)
-}
-
-func mapFuzzyIndex(original, normalized string, normIdx int) int {
-	// Simple approach: count runes up to normIdx in normalized,
-	// then advance that many runes in original.
-	// This works because our normalization only replaces runes 1:1.
-	origRunes := []rune(original)
-	normRunes := []rune(normalized)
-
-	if normIdx >= len(normRunes) {
-		return len(original)
-	}
-
-	// Count bytes for the first normIdx runes in original
-	byteCount := 0
-	for i := 0; i < normIdx && i < len(origRunes); i++ {
-		byteCount += len(string(origRunes[i]))
-	}
-	return byteCount
-}
-
-// generateDiff creates a simple unified diff showing the change.
-func generateDiff(path, old, new string, changeIdx int) string {
-	oldLines := strings.Split(old, "\n")
-	newLines := strings.Split(new, "\n")
-
-	// Find the line number where the change starts
-	lineNum := strings.Count(old[:changeIdx], "\n") + 1
-
-	// Show context around the change
-	contextLines := 3
-	start := max(lineNum-contextLines-1, 0)
-
-	var diff strings.Builder
-	fmt.Fprintf(&diff, "--- %s\n+++ %s\n", path, path)
-
-	// Find changed region
-	endOld := min(lineNum+contextLines+countNewlines(old[changeIdx:])+1, len(oldLines))
-	endNew := min(lineNum+contextLines+countNewlines(new[changeIdx:])+1, len(newLines))
-
-	fmt.Fprintf(&diff, "@@ -%d,%d +%d,%d @@\n", start+1, endOld-start, start+1, endNew-start)
-
-	// Very simplified diff: show old lines as removed, new lines as added
-	// around the change region
-	for i := start; i < endOld && i < len(oldLines); i++ {
-		prefix := " "
-		if i >= lineNum-1 && i < lineNum-1+countNewlines(old[changeIdx:])+1 {
-			prefix = "-"
+	origPos := 0
+	for li, line := range lines {
+		if li > 0 {
+			result = append(result, '\n')
+			mapping = append(mapping, origPos)
+			origPos++ // skip \n in original
 		}
-		fmt.Fprintf(&diff, "%s %s\n", prefix, oldLines[i])
+
+		trimmed := strings.TrimRightFunc(line, unicode.IsSpace)
+
+		for j := 0; j < len(trimmed); {
+			r, size := utf8.DecodeRuneInString(trimmed[j:])
+			repl := normalizeRune(r)
+			for k := 0; k < len(repl); k++ {
+				mapping = append(mapping, origPos+j)
+			}
+			result = append(result, repl...)
+			j += size
+		}
+
+		origPos += len(line) // advance past full original line including trailing ws
 	}
 
-	return diff.String()
+	return string(result), mapping
 }
 
-func countNewlines(s string) int {
-	return strings.Count(s, "\n")
+// normalizeRune maps unicode quotes, dashes, and non-breaking spaces to
+// their ASCII equivalents. Returns the original rune as a string for all
+// other characters.
+func normalizeRune(r rune) string {
+	switch r {
+	case '\u201c', '\u201d': // left/right double quote
+		return "\""
+	case '\u2018', '\u2019': // left/right single quote
+		return "'"
+	case '\u2013', '\u2014': // en dash, em dash
+		return "-"
+	case '\u00a0': // non-breaking space
+		return " "
+	default:
+		return string(r)
+	}
+}
+
+// normalizeForFuzzy normalizes s for fuzzy matching (without position mapping).
+// Used for the search string where position mapping is not needed.
+func normalizeForFuzzy(s string) string {
+	norm, _ := normalizeWithMap(s)
+	return norm
+}
+
+// generateDiff creates a unified diff showing the change between old and new
+// file contents. Uses the go-udiff library for correct diff computation.
+func generateDiff(path, old, new string) string {
+	return udiff.Unified(path, path, old, new)
 }

internal/core/edit_test.go

@@ -0,0 +1,717 @@
+package core
+
+import (
+	"encoding/json"
+	"fmt"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"charm.land/fantasy"
+)
+
+func writeFileOrFail(t *testing.T, path, content string) {
+	t.Helper()
+	if err := os.WriteFile(path, []byte(content), 0644); err != nil {
+		t.Fatalf("failed to write test file %s: %v", path, err)
+	}
+}
+
+// ---------------------------------------------------------------------------
+// fuzzyMatch — the core bug fix
+// ---------------------------------------------------------------------------
+
+func TestFuzzyMatch_TrailingWhitespace(t *testing.T) {
+	// The original bug: trailing whitespace on lines caused mapFuzzyIndex
+	// to return wrong byte positions, corrupting the replacement splice.
+	content := "line1   \nline2   \nline3   \nTAIL\n"
+	search := "line2\nline3"
+
+	idx, matchLen := fuzzyMatch(content, search)
+	if idx < 0 {
+		t.Fatal("expected fuzzy match, got none")
+	}
+
+	matched := content[idx : idx+matchLen]
+	want := "line2   \nline3   "
+	if matched != want {
+		t.Errorf("matched=%q, want=%q", matched, want)
+	}
+
+	// Verify replacement is correct
+	repl := content[:idx] + "REPLACED" + content[idx+matchLen:]
+	wantRepl := "line1   \nREPLACED\nTAIL\n"
+	if repl != wantRepl {
+		t.Errorf("replacement=%q, want=%q", repl, wantRepl)
+	}
+}
+
+func TestFuzzyMatch_TrailingWhitespace_FirstLine(t *testing.T) {
+	content := "line1   \nline2   \nline3\n"
+	search := "line1\nline2"
+
+	idx, matchLen := fuzzyMatch(content, search)
+	if idx < 0 {
+		t.Fatal("expected fuzzy match")
+	}
+
+	matched := content[idx : idx+matchLen]
+	want := "line1   \nline2   "
+	if matched != want {
+		t.Errorf("matched=%q, want=%q", matched, want)
+	}
+}
+
+func TestFuzzyMatch_TrailingWhitespace_LastLine(t *testing.T) {
+	content := "HEAD\nline1   \nline2   \n"
+	search := "line1\nline2"
+
+	idx, matchLen := fuzzyMatch(content, search)
+	if idx < 0 {
+		t.Fatal("expected fuzzy match")
+	}
+
+	matched := content[idx : idx+matchLen]
+	want := "line1   \nline2   "
+	if matched != want {
+		t.Errorf("matched=%q, want=%q", matched, want)
+	}
+}
+
+func TestFuzzyMatch_TrailingWhitespace_AtEOF(t *testing.T) {
+	// Match extends to the very end of the content
+	content := "HEAD\nline1   \nline2   "
+	search := "line1\nline2"
+
+	idx, matchLen := fuzzyMatch(content, search)
+	if idx < 0 {
+		t.Fatal("expected fuzzy match")
+	}
+
+	matched := content[idx : idx+matchLen]
+	want := "line1   \nline2   "
+	if matched != want {
+		t.Errorf("matched=%q, want=%q", matched, want)
+	}
+}
+
+func TestFuzzyMatch_UnicodeQuotes(t *testing.T) {
+	content := "say \u201chello\u201d\n"
+	search := "say \"hello\"\n"
+
+	idx, matchLen := fuzzyMatch(content, search)
+	if idx < 0 {
+		t.Fatal("expected fuzzy match for unicode quotes")
+	}
+
+	matched := content[idx : idx+matchLen]
+	if matched != content { // entire content should match
+		t.Errorf("matched=%q, want=%q", matched, content)
+	}
+}
+
+func TestFuzzyMatch_SmartSingleQuotes(t *testing.T) {
+	content := "it\u2019s a test\n"
+	search := "it's a test\n"
+
+	idx, matchLen := fuzzyMatch(content, search)
+	if idx < 0 {
+		t.Fatal("expected fuzzy match for smart single quotes")
+	}
+	matched := content[idx : idx+matchLen]
+	if matched != content {
+		t.Errorf("matched=%q, want=%q", matched, content)
+	}
+}
+
+func TestFuzzyMatch_EmDash(t *testing.T) {
+	content := "foo \u2014 bar\n"
+	search := "foo - bar\n"
+
+	idx, matchLen := fuzzyMatch(content, search)
+	if idx < 0 {
+		t.Fatal("expected fuzzy match for em dash")
+	}
+	matched := content[idx : idx+matchLen]
+	if matched != content {
+		t.Errorf("matched=%q, want=%q", matched, content)
+	}
+}
+
+func TestFuzzyMatch_NonBreakingSpace(t *testing.T) {
+	content := "hello\u00a0world\n"
+	search := "hello world\n"
+
+	idx, matchLen := fuzzyMatch(content, search)
+	if idx < 0 {
+		t.Fatal("expected fuzzy match for non-breaking space")
+	}
+	matched := content[idx : idx+matchLen]
+	if matched != content {
+		t.Errorf("matched=%q, want=%q", matched, content)
+	}
+}
+
+func TestFuzzyMatch_NoMatch(t *testing.T) {
+	content := "hello world\n"
+	search := "goodbye world\n"
+
+	idx, _ := fuzzyMatch(content, search)
+	if idx >= 0 {
+		t.Error("expected no match")
+	}
+}
+
+func TestFuzzyMatch_AmbiguousReturnsNoMatch(t *testing.T) {
+	// Two identical blocks — fuzzy match should refuse to pick one
+	content := "block\nblock\n"
+	search := "block"
+
+	idx, _ := fuzzyMatch(content, search)
+	if idx >= 0 {
+		t.Error("expected no match for ambiguous fuzzy hit")
+	}
+}
+
+func TestFuzzyMatch_EmptySearch(t *testing.T) {
+	idx, _ := fuzzyMatch("content", "")
+	if idx >= 0 {
+		t.Error("expected no match for empty search")
+	}
+}
+
+func TestFuzzyMatch_MultiLineWithMixedWhitespace(t *testing.T) {
+	content := "func foo() {\t  \n\treturn 1  \n}\t \n"
+	search := "func foo() {\n\treturn 1\n}"
+
+	idx, matchLen := fuzzyMatch(content, search)
+	if idx < 0 {
+		t.Fatal("expected fuzzy match")
+	}
+
+	// Replacement should preserve surrounding content
+	repl := content[:idx] + "func bar() {\n\treturn 2\n}" + content[idx+matchLen:]
+	if !strings.HasPrefix(repl, "func bar()") {
+		t.Errorf("unexpected replacement start: %q", repl[:20])
+	}
+	if !strings.HasSuffix(repl, "\n") {
+		t.Errorf("replacement should end with newline: %q", repl)
+	}
+}
+
+// ---------------------------------------------------------------------------
+// normalizeWithMap — position mapping correctness
+// ---------------------------------------------------------------------------
+
+func TestNormalizeWithMap_NoTrailingWhitespace(t *testing.T) {
+	s := "abc\ndef"
+	norm, mapping := normalizeWithMap(s)
+	if norm != s {
+		t.Errorf("norm=%q, want=%q", norm, s)
+	}
+	// Each byte should map to itself
+	for i, orig := range mapping {
+		if orig != i {
+			t.Errorf("mapping[%d]=%d, want=%d", i, orig, i)
+		}
+	}
+}
+
+func TestNormalizeWithMap_TrailingWhitespace(t *testing.T) {
+	s := "ab   \ncd"
+	norm, mapping := normalizeWithMap(s)
+	wantNorm := "ab\ncd"
+	if norm != wantNorm {
+		t.Errorf("norm=%q, want=%q", norm, wantNorm)
+	}
+	// 'a'→0, 'b'→1, '\n'→5, 'c'→6, 'd'→7
+	wantMapping := []int{0, 1, 5, 6, 7}
+	if len(mapping) != len(wantMapping) {
+		t.Fatalf("mapping len=%d, want=%d", len(mapping), len(wantMapping))
+	}
+	for i, want := range wantMapping {
+		if mapping[i] != want {
+			t.Errorf("mapping[%d]=%d, want=%d", i, mapping[i], want)
+		}
+	}
+}
+
+func TestNormalizeWithMap_UnicodeReplacement(t *testing.T) {
+	// \u201c is 3 bytes in UTF-8, replaced with " which is 1 byte
+	s := "\u201chello\u201d"
+	norm, mapping := normalizeWithMap(s)
+	wantNorm := "\"hello\""
+	if norm != wantNorm {
+		t.Errorf("norm=%q, want=%q", norm, wantNorm)
+	}
+	// " maps to byte 0 (start of \u201c), h maps to 3, e→4, l→5, l→6, o→7, " maps to 8 (start of \u201d)
+	wantMapping := []int{0, 3, 4, 5, 6, 7, 8}
+	if len(mapping) != len(wantMapping) {
+		t.Fatalf("mapping len=%d, want=%d", len(mapping), len(wantMapping))
+	}
+	for i, want := range wantMapping {
+		if mapping[i] != want {
+			t.Errorf("mapping[%d]=%d, want=%d", i, mapping[i], want)
+		}
+	}
+}
+
+func TestNormalizeWithMap_EmptyString(t *testing.T) {
+	norm, mapping := normalizeWithMap("")
+	if norm != "" {
+		t.Errorf("norm=%q, want empty", norm)
+	}
+	if len(mapping) != 0 {
+		t.Errorf("mapping len=%d, want 0", len(mapping))
+	}
+}
+
+func TestNormalizeWithMap_OnlyWhitespace(t *testing.T) {
+	norm, _ := normalizeWithMap("   \n   ")
+	if norm != "\n" {
+		t.Errorf("norm=%q, want %q", norm, "\n")
+	}
+}
+
+// ---------------------------------------------------------------------------
+// normalizeForFuzzy — consistency with normalizeWithMap
+// ---------------------------------------------------------------------------
+
+func TestNormalizeForFuzzy_ConsistentWithMap(t *testing.T) {
+	inputs := []string{
+		"hello   \nworld   ",
+		"\u201chello\u201d\u2014world",
+		"a\u00a0b\u2013c\n  trailing  \n",
+		"no changes here",
+		"",
+	}
+	for _, s := range inputs {
+		norm := normalizeForFuzzy(s)
+		normMap, _ := normalizeWithMap(s)
+		if norm != normMap {
+			t.Errorf("normalizeForFuzzy(%q) = %q, normalizeWithMap = %q", s, norm, normMap)
+		}
+	}
+}
+
+// ---------------------------------------------------------------------------
+// generateDiff — correct unified diff output
+// ---------------------------------------------------------------------------
+
+func TestGenerateDiff_SingleLineChange(t *testing.T) {
+	old := "line1\nline2\nline3\nline4\nline5\nline6\nline7\n"
+	new := "line1\nline2\nline3\nLINE4\nline5\nline6\nline7\n"
+
+	diff := generateDiff("test.go", old, new)
+
+	// Should contain standard unified diff markers
+	if !strings.Contains(diff, "--- test.go") {
+		t.Error("diff should contain --- header")
+	}
+	if !strings.Contains(diff, "+++ test.go") {
+		t.Error("diff should contain +++ header")
+	}
+	if !strings.Contains(diff, "@@") {
+		t.Error("diff should contain @@ hunk header")
+	}
+
+	// Should show the actual change
+	if !strings.Contains(diff, "-line4") {
+		t.Error("diff should show removed line")
+	}
+	if !strings.Contains(diff, "+LINE4") {
+		t.Error("diff should show added line")
+	}
+
+	// Should NOT mark all remaining lines as changed (the old bug)
+	deletedCount := strings.Count(diff, "\n-")
+	if deletedCount > 2 { // at most 1 deleted line + some tolerance
+		t.Errorf("diff shows %d deletions, expected ~1 (old bug: marked rest of file as deleted)", deletedCount)
+	}
+}
+
+func TestGenerateDiff_MultiLineChange(t *testing.T) {
+	old := "aaa\nbbb\nccc\nddd\n"
+	new := "aaa\nBBB\nCCC\nddd\n"
+
+	diff := generateDiff("x.go", old, new)
+	if !strings.Contains(diff, "-bbb") {
+		t.Error("diff should show bbb removed")
+	}
+	if !strings.Contains(diff, "-ccc") {
+		t.Error("diff should show ccc removed")
+	}
+	if !strings.Contains(diff, "+BBB") {
+		t.Error("diff should show BBB added")
+	}
+	if !strings.Contains(diff, "+CCC") {
+		t.Error("diff should show CCC added")
+	}
+}
+
+func TestGenerateDiff_NoChange(t *testing.T) {
+	content := "hello\nworld\n"
+	diff := generateDiff("x.go", content, content)
+	if diff != "" {
+		t.Errorf("expected empty diff for identical content, got %q", diff)
+	}
+}
+
+func TestGenerateDiff_Addition(t *testing.T) {
+	old := "line1\nline2\n"
+	new := "line1\nnew line\nline2\n"
+
+	diff := generateDiff("x.go", old, new)
+	if !strings.Contains(diff, "+new line") {
+		t.Error("diff should show added line")
+	}
+}
+
+func TestGenerateDiff_Deletion(t *testing.T) {
+	old := "line1\nremove me\nline2\n"
+	new := "line1\nline2\n"
+
+	diff := generateDiff("x.go", old, new)
+	if !strings.Contains(diff, "-remove me") {
+		t.Error("diff should show deleted line")
+	}
+}
+
+// ---------------------------------------------------------------------------
+// End-to-end: executeEdit via tool call
+// ---------------------------------------------------------------------------
+
+func TestExecuteEdit_ExactMatch(t *testing.T) {
+	dir := t.TempDir()
+	path := filepath.Join(dir, "test.go")
+	original := "func main() {\n\tfmt.Println(\"hello\")\n}\n"
+	writeFileOrFail(t, path, original)
+
+	input, _ := json.Marshal(editArgs{
+		Path:    path,
+		OldText: "fmt.Println(\"hello\")",
+		NewText: "fmt.Println(\"world\")",
+	})
+
+	resp, err := executeEdit(t.Context(), fantasy.ToolCall{Input: string(input)}, dir)
+	if err != nil {
+		t.Fatalf("executeEdit error: %v", err)
+	}
+	if resp.IsError {
+		t.Fatalf("tool returned error: %s", resp.Content)
+	}
+
+	got, _ := os.ReadFile(path)
+	want := "func main() {\n\tfmt.Println(\"world\")\n}\n"
+	if string(got) != want {
+		t.Errorf("file content=%q, want=%q", string(got), want)
+	}
+}
+
+func TestExecuteEdit_ExactMatch_DoesNotCorruptRest(t *testing.T) {
+	// This is the key regression test for the screenshot bug: editing a
+	// small section must NOT delete/corrupt the rest of the file.
+	dir := t.TempDir()
+	path := filepath.Join(dir, "big.go")
+
+	var lines []string
+	for i := 1; i <= 100; i++ {
+		lines = append(lines, fmt.Sprintf("line_%03d_%s", i, strings.Repeat("x", 40)))
+	}
+	original := strings.Join(lines, "\n") + "\n"
+	writeFileOrFail(t, path, original)
+
+	// Replace just line 50
+	target := lines[49]
+	replacement := "REPLACED_LINE_50"
+	input, _ := json.Marshal(editArgs{
+		Path:    path,
+		OldText: target,
+		NewText: replacement,
+	})
+
+	resp, err := executeEdit(t.Context(), fantasy.ToolCall{Input: string(input)}, dir)
+	if err != nil {
+		t.Fatalf("executeEdit error: %v", err)
+	}
+	if resp.IsError {
+		t.Fatalf("tool returned error: %s", resp.Content)
+	}
+
+	got, _ := os.ReadFile(path)
+	gotLines := strings.Split(string(got), "\n")
+
+	// File should still have 101 elements (100 lines + trailing empty)
+	if len(gotLines) != 101 {
+		t.Fatalf("file has %d lines, want 101 (content was corrupted)", len(gotLines))
+	}
+
+	// Line 50 should be replaced
+	if gotLines[49] != replacement {
+		t.Errorf("line 50=%q, want=%q", gotLines[49], replacement)
+	}
+
+	// Lines before and after should be untouched
+	if gotLines[0] != lines[0] {
+		t.Errorf("line 1 corrupted: %q", gotLines[0])
+	}
+	if gotLines[98] != lines[98] {
+		t.Errorf("line 99 corrupted: %q", gotLines[98])
+	}
+}
+
+func TestExecuteEdit_FuzzyMatch_TrailingWhitespace(t *testing.T) {
+	dir := t.TempDir()
+	path := filepath.Join(dir, "ws.go")
+	// File has trailing whitespace on some lines
+	original := "func foo() {   \n\treturn 1   \n}\nfunc bar() {\n}\n"
+	writeFileOrFail(t, path, original)
+
+	// Search without trailing whitespace (common LLM behavior)
+	input, _ := json.Marshal(editArgs{
+		Path:    path,
+		OldText: "func foo() {\n\treturn 1\n}",
+		NewText: "func foo() {\n\treturn 2\n}",
+	})
+
+	resp, err := executeEdit(t.Context(), fantasy.ToolCall{Input: string(input)}, dir)
+	if err != nil {
+		t.Fatalf("executeEdit error: %v", err)
+	}
+	if resp.IsError {
+		t.Fatalf("tool returned error: %s", resp.Content)
+	}
+
+	got, _ := os.ReadFile(path)
+	gotStr := string(got)
+
+	// The fuzzy match replaces the matched region (which includes trailing
+	// whitespace) with the new_text. The key invariant is that the rest of
+	// the file (func bar) must be preserved.
+	if !strings.Contains(gotStr, "return 2") {
+		t.Error("edit was not applied: missing 'return 2'")
+	}
+	if !strings.Contains(gotStr, "func bar()") {
+		t.Errorf("file was corrupted: missing func bar(). got=%q", gotStr)
+	}
+
+	// Verify response mentions fuzzy match
+	if !strings.Contains(resp.Content, "fuzzy match") {
+		t.Error("response should mention fuzzy match")
+	}
+}
+
+func TestExecuteEdit_FuzzyMatch_DoesNotCorruptRest(t *testing.T) {
+	// Regression test: fuzzy match must not corrupt content after the match.
+	dir := t.TempDir()
+	path := filepath.Join(dir, "fuzzy.txt")
+
+	// 20 lines, each with trailing whitespace
+	var lines []string
+	for i := 1; i <= 20; i++ {
+		lines = append(lines, strings.Repeat("x", 10)+"   ") // trailing spaces
+	}
+	original := strings.Join(lines, "\n") + "\nEND\n"
+	writeFileOrFail(t, path, original)
+
+	// Search for lines 10-11 without trailing whitespace
+	search := strings.Repeat("x", 10) + "\n" + strings.Repeat("x", 10)
+	// But this matches lines 1-2, 2-3, etc. — should fail due to ambiguity.
+	input, _ := json.Marshal(editArgs{
+		Path:    path,
+		OldText: search,
+		NewText: "REPLACED",
+	})
+
+	resp, err := executeEdit(t.Context(), fantasy.ToolCall{Input: string(input)}, dir)
+	if err != nil {
+		t.Fatalf("executeEdit error: %v", err)
+	}
+
+	// This should either fail (ambiguous) or produce correct output.
+	// With identical lines, fuzzy match should refuse (ambiguous).
+	got, _ := os.ReadFile(path)
+	if !resp.IsError {
+		// If it didn't error, verify the file is not corrupted
+		if !strings.HasSuffix(string(got), "END\n") {
+			t.Error("file was corrupted: missing END marker")
+		}
+	}
+}
+
+func TestExecuteEdit_MultipleMatches_Fails(t *testing.T) {
+	dir := t.TempDir()
+	path := filepath.Join(dir, "dup.txt")
+	writeFileOrFail(t, path, "hello\nworld\nhello\n")
+
+	input, _ := json.Marshal(editArgs{
+		Path:    path,
+		OldText: "hello",
+		NewText: "goodbye",
+	})
+
+	resp, err := executeEdit(t.Context(), fantasy.ToolCall{Input: string(input)}, dir)
+	if err != nil {
+		t.Fatalf("executeEdit error: %v", err)
+	}
+	if !resp.IsError {
+		t.Error("expected error for multiple matches")
+	}
+	if !strings.Contains(resp.Content, "2 matches") {
+		t.Errorf("expected '2 matches' in error, got: %s", resp.Content)
+	}
+
+	// File should be untouched
+	got, _ := os.ReadFile(path)
+	if string(got) != "hello\nworld\nhello\n" {
+		t.Error("file was modified despite error")
+	}
+}
+
+func TestExecuteEdit_NoMatch_Fails(t *testing.T) {
+	dir := t.TempDir()
+	path := filepath.Join(dir, "nomatch.txt")
+	writeFileOrFail(t, path, "hello world\n")
+
+	input, _ := json.Marshal(editArgs{
+		Path:    path,
+		OldText: "nonexistent text",
+		NewText: "replacement",
+	})
+
+	resp, err := executeEdit(t.Context(), fantasy.ToolCall{Input: string(input)}, dir)
+	if err != nil {
+		t.Fatalf("executeEdit error: %v", err)
+	}
+	if !resp.IsError {
+		t.Error("expected error for no match")
+	}
+
+	// File should be untouched
+	got, _ := os.ReadFile(path)
+	if string(got) != "hello world\n" {
+		t.Error("file was modified despite error")
+	}
+}
+
+func TestExecuteEdit_CRLFNormalization(t *testing.T) {
+	dir := t.TempDir()
+	path := filepath.Join(dir, "crlf.txt")
+	writeFileOrFail(t, path, "line1\r\nline2\r\nline3\r\n")
+
+	input, _ := json.Marshal(editArgs{
+		Path:    path,
+		OldText: "line2",
+		NewText: "LINE2",
+	})
+
+	resp, err := executeEdit(t.Context(), fantasy.ToolCall{Input: string(input)}, dir)
+	if err != nil {
+		t.Fatalf("executeEdit error: %v", err)
+	}
+	if resp.IsError {
+		t.Fatalf("tool returned error: %s", resp.Content)
+	}
+
+	got, _ := os.ReadFile(path)
+	if !strings.Contains(string(got), "LINE2") {
+		t.Error("edit was not applied")
+	}
+}
+
+func TestExecuteEdit_MissingPath(t *testing.T) {
+	input, _ := json.Marshal(editArgs{
+		OldText: "x",
+		NewText: "y",
+	})
+	resp, err := executeEdit(t.Context(), fantasy.ToolCall{Input: string(input)}, "")
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if !resp.IsError {
+		t.Error("expected error for missing path")
+	}
+}
+
+func TestExecuteEdit_NonexistentFile(t *testing.T) {
+	input, _ := json.Marshal(editArgs{
+		Path:    "/tmp/nonexistent_edit_test_file_12345.go",
+		OldText: "x",
+		NewText: "y",
+	})
+	resp, err := executeEdit(t.Context(), fantasy.ToolCall{Input: string(input)}, "")
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if !resp.IsError {
+		t.Error("expected error for nonexistent file")
+	}
+}
+
+func TestExecuteEdit_DiffContainsHunkHeader(t *testing.T) {
+	// The UI's extractDiffStartLine parses @@ -N from the result.
+	// Verify the diff output contains it.
+	dir := t.TempDir()
+	path := filepath.Join(dir, "hunk.go")
+	var lines []string
+	for i := 1; i <= 20; i++ {
+		lines = append(lines, fmt.Sprintf("line_%02d_content", i))
+	}
+	writeFileOrFail(t, path, strings.Join(lines, "\n")+"\n")
+
+	input, _ := json.Marshal(editArgs{
+		Path:    path,
+		OldText: "line_10_content",
+		NewText: "REPLACED",
+	})
+
+	resp, err := executeEdit(t.Context(), fantasy.ToolCall{Input: string(input)}, dir)
+	if err != nil {
+		t.Fatalf("error: %v", err)
+	}
+	if resp.IsError {
+		t.Fatalf("tool error: %s", resp.Content)
+	}
+	if !strings.Contains(resp.Content, "@@ ") {
+		t.Error("diff output should contain @@ hunk header for UI parsing")
+	}
+}
+
+func TestExecuteEdit_MetadataContainsFileDiffs(t *testing.T) {
+	dir := t.TempDir()
+	path := filepath.Join(dir, "meta.go")
+	writeFileOrFail(t, path, "old content\n")
+
+	input, _ := json.Marshal(editArgs{
+		Path:    path,
+		OldText: "old content",
+		NewText: "new content",
+	})
+
+	resp, err := executeEdit(t.Context(), fantasy.ToolCall{Input: string(input)}, dir)
+	if err != nil {
+		t.Fatalf("error: %v", err)
+	}
+
+	// Check metadata is present
+	metaJSON := resp.Metadata
+	if metaJSON == "" {
+		t.Fatal("expected metadata on response")
+	}
+
+	var meta map[string]any
+	if err := json.Unmarshal([]byte(metaJSON), &meta); err != nil {
+		t.Fatalf("metadata is not valid JSON: %v", err)
+	}
+
+	diffs, ok := meta["file_diffs"]
+	if !ok {
+		t.Fatal("metadata missing file_diffs key")
+	}
+
+	diffList, ok := diffs.([]any)
+	if !ok || len(diffList) == 0 {
+		t.Fatal("file_diffs should be a non-empty array")
+	}
+}

internal/core/subagent.go

@@ -28,7 +28,9 @@ type SubagentSpawnResult struct {
 // SubagentSpawnFunc is a callback that spawns an in-process subagent. The
 // parent Kit instance injects this into the context so the core tool can
 // call back without importing pkg/kit (which would create a cycle).
-type SubagentSpawnFunc func(ctx context.Context, prompt, model, systemPrompt string, timeout time.Duration) (*SubagentSpawnResult, error)
+// The toolCallID parameter is the LLM-assigned ID of the spawn_subagent
+// tool call, enabling the parent to correlate subagent events.
+type SubagentSpawnFunc func(ctx context.Context, toolCallID, prompt, model, systemPrompt string, timeout time.Duration) (*SubagentSpawnResult, error)
 
 type subagentCtxKey struct{}
 
@@ -128,8 +130,16 @@ func executeSubagent(ctx context.Context, call fantasy.ToolCall) (fantasy.ToolRe
 		), fmt.Errorf("no subagent spawner in context")
 	}
 
+	// Detach from the parent's deadline so the subagent gets its own
+	// independent timeout (applied downstream in Kit.Subagent). The parent
+	// context may carry a tight deadline from the LLM generation loop or
+	// other tool timeouts that would prematurely kill the subagent.
+	// We preserve context values (spawner, etc.) and propagate parent
+	// cancellation (e.g. user hits Ctrl-C) without inheriting the deadline.
+	spawnCtx := detachedWithCancel(ctx)
+
 	// Spawn in-process subagent.
-	result, err := spawner(ctx, args.Task, args.Model, args.SystemPrompt, timeout)
+	result, err := spawner(spawnCtx, call.ID, args.Task, args.Model, args.SystemPrompt, timeout)
 	if err != nil || result.Error != nil {
 		spawnErr := err
 		if spawnErr == nil {
@@ -162,6 +172,39 @@ func executeSubagent(ctx context.Context, call fantasy.ToolCall) (fantasy.ToolRe
 	return resp, nil
 }
 
+// ---------------------------------------------------------------------------
+// Context detachment
+// ---------------------------------------------------------------------------
+
+// detachedContext wraps a parent context, preserving its values but removing
+// its deadline and cancellation. This allows the subagent to have its own
+// independent timeout while still accessing context-stored values (e.g. the
+// subagent spawner function).
+type detachedContext struct {
+	parent context.Context
+}
+
+func (d detachedContext) Deadline() (time.Time, bool) { return time.Time{}, false }
+func (d detachedContext) Done() <-chan struct{}       { return nil }
+func (d detachedContext) Err() error                  { return nil }
+func (d detachedContext) Value(key any) any           { return d.parent.Value(key) }
+
+// detachedWithCancel creates a new context that inherits values from the
+// parent but has no deadline. Cancellation of the parent is propagated: when
+// the parent is cancelled the returned context is also cancelled, but the
+// parent's deadline does not apply to the child.
+func detachedWithCancel(parent context.Context) context.Context {
+	child, cancel := context.WithCancel(detachedContext{parent: parent})
+	go func() {
+		select {
+		case <-parent.Done():
+			cancel()
+		case <-child.Done():
+		}
+	}()
+	return child
+}
+
 // truncateResponse limits the response length to avoid overwhelming context windows.
 func truncateResponse(s string, maxLen int) string {
 	if len(s) <= maxLen {

internal/core/truncate.go

@@ -6,14 +6,17 @@ import (
 )
 
 const (
-	defaultMaxLines = 2000
-	defaultMaxBytes = 50 * 1024 // 50KB
-	grepMaxLineLen  = 500
+	defaultMaxLines   = 2000
+	defaultMaxBytes   = 50 * 1024 // 50KB
+	defaultMaxLineLen = 2000      // max characters per line before truncation
+	grepMaxLineLen    = 500
 
 	// DefaultMaxLines is the exported default line limit for truncation.
 	DefaultMaxLines = defaultMaxLines
 	// DefaultMaxBytes is the exported default byte limit for truncation.
 	DefaultMaxBytes = defaultMaxBytes
+	// DefaultMaxLineLen is the exported default per-line character limit.
+	DefaultMaxLineLen = defaultMaxLineLen
 )
 
 // TruncationResult describes how output was truncated.
@@ -26,6 +29,8 @@ type TruncationResult struct {
 }
 
 // TruncateTail keeps the last maxLines lines and at most maxBytes bytes.
+// Individual lines longer than defaultMaxLineLen are truncated to prevent
+// extremely long single lines from blowing up the TUI when wrapped.
 // Used for bash output where the tail is most relevant.
 func TruncateTail(content string, maxLines, maxBytes int) TruncationResult {
 	if maxLines <= 0 {
@@ -38,11 +43,11 @@ func TruncateTail(content string, maxLines, maxBytes int) TruncationResult {
 	lines := strings.Split(content, "\n")
 	total := len(lines)
 
-	if len(content) <= maxBytes && total <= maxLines {
-		return TruncationResult{Content: content, Total: total, Kept: total}
-	}
+	// Truncate individual long lines first to prevent single lines from
+	// wrapping into hundreds of visual lines in the TUI.
+	lines = truncateLongLines(lines, defaultMaxLineLen)
 
-	// Truncate by lines first (keep tail)
+	// Truncate by lines (keep tail)
 	truncBy := ""
 	if total > maxLines {
 		lines = lines[total-maxLines:]
@@ -78,6 +83,7 @@ func TruncateTail(content string, maxLines, maxBytes int) TruncationResult {
 }
 
 // truncateHead keeps the first maxLines lines and at most maxBytes bytes.
+// Individual lines longer than defaultMaxLineLen are truncated.
 // Used for read, grep, find, ls output where the head is most relevant.
 func truncateHead(content string, maxLines, maxBytes int) TruncationResult {
 	if maxLines <= 0 {
@@ -90,9 +96,8 @@ func truncateHead(content string, maxLines, maxBytes int) TruncationResult {
 	lines := strings.Split(content, "\n")
 	total := len(lines)
 
-	if len(content) <= maxBytes && total <= maxLines {
-		return TruncationResult{Content: content, Total: total, Kept: total}
-	}
+	// Truncate individual long lines first.
+	lines = truncateLongLines(lines, defaultMaxLineLen)
 
 	truncBy := ""
 	if total > maxLines {
@@ -125,6 +130,19 @@ func truncateHead(content string, maxLines, maxBytes int) TruncationResult {
 	}
 }
 
+// truncateLongLines caps each line to maxLen characters, appending a
+// "[...N chars truncated]" marker to any line that exceeds the limit.
+// This prevents a single very long line (e.g. minified JSON/JS) from
+// wrapping into hundreds of visual rows and blowing up the TUI.
+func truncateLongLines(lines []string, maxLen int) []string {
+	for i, line := range lines {
+		if len(line) > maxLen {
+			lines[i] = line[:maxLen] + fmt.Sprintf("... [%d chars truncated]", len(line)-maxLen)
+		}
+	}
+	return lines
+}
+
 // truncateLine truncates a single line to maxChars, appending "..." if cut.
 func truncateLine(line string, maxChars int) string {
 	if maxChars <= 0 {

internal/core/truncate_test.go

@@ -0,0 +1,163 @@
+package core
+
+import (
+	"strings"
+	"testing"
+)
+
+func TestTruncateTail_LongLines(t *testing.T) {
+	// A single line of 5000 chars should be truncated to defaultMaxLineLen.
+	longLine := strings.Repeat("x", 5000)
+	tr := TruncateTail(longLine, 2000, 50*1024)
+
+	if len(tr.Content) > defaultMaxLineLen+100 { // +100 for the "[...N chars truncated]" suffix
+		t.Errorf("single long line not truncated: got %d chars, want <= %d", len(tr.Content), defaultMaxLineLen+100)
+	}
+	if !strings.Contains(tr.Content, "chars truncated]") {
+		t.Error("truncated line should contain truncation marker")
+	}
+}
+
+func TestTruncateTail_NormalLines(t *testing.T) {
+	// Lines within the limit should pass through unchanged.
+	content := "line1\nline2\nline3"
+	tr := TruncateTail(content, 2000, 50*1024)
+	if tr.Content != content {
+		t.Errorf("got %q, want %q", tr.Content, content)
+	}
+	if tr.Truncated {
+		t.Error("should not be marked as truncated")
+	}
+}
+
+func TestTruncateTail_LineCount(t *testing.T) {
+	lines := make([]string, 100)
+	for i := range lines {
+		lines[i] = "line"
+	}
+	content := strings.Join(lines, "\n")
+	tr := TruncateTail(content, 10, 50*1024)
+
+	if !tr.Truncated {
+		t.Error("should be marked as truncated")
+	}
+	if tr.Total != 100 {
+		t.Errorf("total = %d, want 100", tr.Total)
+	}
+	if tr.Kept != 10 {
+		t.Errorf("kept = %d, want 10", tr.Kept)
+	}
+}
+
+func TestTruncateHead_LongLines(t *testing.T) {
+	longLine := strings.Repeat("y", 5000)
+	tr := truncateHead(longLine, 2000, 50*1024)
+
+	if len(tr.Content) > defaultMaxLineLen+100 {
+		t.Errorf("single long line not truncated: got %d chars, want <= %d", len(tr.Content), defaultMaxLineLen+100)
+	}
+	if !strings.Contains(tr.Content, "chars truncated]") {
+		t.Error("truncated line should contain truncation marker")
+	}
+}
+
+func TestTruncateHead_NormalLines(t *testing.T) {
+	content := "line1\nline2\nline3"
+	tr := truncateHead(content, 2000, 50*1024)
+	if tr.Content != content {
+		t.Errorf("got %q, want %q", tr.Content, content)
+	}
+	if tr.Truncated {
+		t.Error("should not be marked as truncated")
+	}
+}
+
+func TestTruncateHead_LineCount(t *testing.T) {
+	lines := make([]string, 100)
+	for i := range lines {
+		lines[i] = "line"
+	}
+	content := strings.Join(lines, "\n")
+	tr := truncateHead(content, 10, 50*1024)
+
+	if !tr.Truncated {
+		t.Error("should be marked as truncated")
+	}
+	if tr.Total != 100 {
+		t.Errorf("total = %d, want 100", tr.Total)
+	}
+	if tr.Kept != 10 {
+		t.Errorf("kept = %d, want 10", tr.Kept)
+	}
+}
+
+func TestTruncateLongLines(t *testing.T) {
+	lines := []string{
+		"short",
+		strings.Repeat("a", 3000),
+		"also short",
+	}
+	result := truncateLongLines(lines, 100)
+
+	if result[0] != "short" {
+		t.Error("short line should be unchanged")
+	}
+	if len(result[1]) > 200 { // 100 chars + marker
+		t.Errorf("long line not truncated: len=%d", len(result[1]))
+	}
+	if !strings.Contains(result[1], "chars truncated]") {
+		t.Error("should contain truncation marker")
+	}
+	if result[2] != "also short" {
+		t.Error("short line should be unchanged")
+	}
+}
+
+func TestTruncateTail_MixedLongAndManyLines(t *testing.T) {
+	// 50 lines, each 3000 chars — tests both per-line and total truncation.
+	lines := make([]string, 50)
+	for i := range lines {
+		lines[i] = strings.Repeat("z", 3000)
+	}
+	content := strings.Join(lines, "\n")
+
+	tr := TruncateTail(content, 10, 50*1024)
+
+	// Should keep 10 lines.
+	if tr.Kept != 10 {
+		t.Errorf("kept = %d, want 10", tr.Kept)
+	}
+	// Each line should be capped at ~defaultMaxLineLen.
+	resultLines := strings.Split(tr.Content, "\n")
+	for i, line := range resultLines {
+		if len(line) > defaultMaxLineLen+100 {
+			t.Errorf("line %d too long: %d chars", i, len(line))
+		}
+	}
+}
+
+func TestTruncateLine(t *testing.T) {
+	short := "hello"
+	if truncateLine(short, 10) != short {
+		t.Error("short line should be unchanged")
+	}
+
+	long := strings.Repeat("x", 100)
+	result := truncateLine(long, 10)
+	if len(result) != 13 { // 10 + "..."
+		t.Errorf("got len %d, want 13", len(result))
+	}
+
+	// Default max for 0 — input shorter than default, so unchanged
+	result2 := truncateLine(long, 0)
+	if result2 != long {
+		t.Errorf("100-char line should be unchanged when maxChars defaults to %d", grepMaxLineLen)
+	}
+
+	// Longer input with default
+	veryLong := strings.Repeat("x", 1000)
+	result3 := truncateLine(veryLong, 0)
+	if len(result3) != grepMaxLineLen+3 {
+		t.Errorf("got len %d, want %d", len(result3), grepMaxLineLen+3)
+	}
+}

internal/extensions/api.go

@@ -727,6 +727,7 @@ type API struct {
 	onToolCall                func(func(ToolCallEvent, Context) *ToolCallResult)
 	onToolExecStart           func(func(ToolExecutionStartEvent, Context))
 	onToolExecEnd             func(func(ToolExecutionEndEvent, Context))
+	onToolOutput              func(func(ToolOutputEvent, Context))
 	onToolResult              func(func(ToolResultEvent, Context) *ToolResultResult)
 	onInput                   func(func(InputEvent, Context) *InputResult)
 	onBeforeAgentStart        func(func(BeforeAgentStartEvent, Context) *BeforeAgentStartResult)
@@ -767,6 +768,13 @@ func (a *API) OnToolExecutionEnd(handler func(ToolExecutionEndEvent, Context)) {
 	a.onToolExecEnd(handler)
 }
 
+// OnToolOutput registers a handler for streaming tool output chunks.
+// This fires for each output line as it arrives from tools like bash,
+// allowing extensions to observe or process output in real-time.
+func (a *API) OnToolOutput(handler func(ToolOutputEvent, Context)) {
+	a.onToolOutput(handler)
+}
+
 // OnToolResult registers a handler that fires after tool execution.
 // Return a non-nil ToolResultResult to modify the output.
 func (a *API) OnToolResult(handler func(ToolResultEvent, Context) *ToolResultResult) {
@@ -1538,6 +1546,19 @@ type ToolExecutionEndEvent struct {
 
 func (e ToolExecutionEndEvent) Type() EventType { return ToolExecutionEnd }
 
+// ToolOutputEvent fires when a tool produces streaming output chunks.
+// This is primarily used for long-running tools like bash to show output
+// in real-time as it arrives, before the tool completes.
+type ToolOutputEvent struct {
+	ToolCallID string
+	ToolName   string
+	ToolKind   string
+	Chunk      string // Output text chunk
+	IsStderr   bool   // Whether this chunk came from stderr
+}
+
+func (e ToolOutputEvent) Type() EventType { return ToolOutput }
+
 // ToolResultEvent fires after tool execution with the output.
 type ToolResultEvent struct {
 	ToolCallID string
@@ -1743,13 +1764,18 @@ type BeforeCompactEvent struct {
 func (e BeforeCompactEvent) Type() EventType { return BeforeCompact }
 
 // BeforeCompactResult controls whether compaction proceeds. Return
-// Cancel=true with an optional Reason to block compaction.
+// Cancel=true with an optional Reason to block compaction, or provide
+// a custom Summary to replace the default LLM-generated one.
 type BeforeCompactResult struct {
 	// Cancel, when true, prevents compaction from proceeding.
 	Cancel bool
 	// Reason is a human-readable explanation shown to the user when
 	// Cancel is true. Empty string uses a default message.
 	Reason string
+	// Summary, when non-empty, replaces the default LLM-generated summary.
+	// The extension is responsible for generating a useful summary.
+	// Ignored when Cancel is true.
+	Summary string
 }
 
 func (BeforeCompactResult) isResult() {}

internal/extensions/events.go

@@ -19,6 +19,9 @@ const (
 	// ToolExecutionEnd fires when a tool finishes executing.
 	ToolExecutionEnd EventType = "tool_execution_end"
 
+	// ToolOutput fires when a tool produces streaming output chunks.
+	ToolOutput EventType = "tool_output"
+
 	// ToolResult fires after a tool executes. Handlers can modify the result.
 	ToolResult EventType = "tool_result"
 

internal/extensions/loader.go

@@ -439,6 +439,12 @@ func loadSingleExtension(path string) (*LoadedExtension, error) {
 				return nil
 			})
 		},
+		onToolOutput: func(h func(ToolOutputEvent, Context)) {
+			reg(ToolOutput, func(e Event, c Context) Result {
+				h(e.(ToolOutputEvent), c)
+				return nil
+			})
+		},
 		onToolResult: func(h func(ToolResultEvent, Context) *ToolResultResult) {
 			reg(ToolResult, func(e Event, c Context) Result {
 				r := h(e.(ToolResultEvent), c)

internal/extensions/symbols.go

@@ -128,6 +128,7 @@ func Symbols() interp.Exports {
 			"ToolCallResult":          reflect.ValueOf((*ToolCallResult)(nil)),
 			"ToolExecutionStartEvent": reflect.ValueOf((*ToolExecutionStartEvent)(nil)),
 			"ToolExecutionEndEvent":   reflect.ValueOf((*ToolExecutionEndEvent)(nil)),
+			"ToolOutputEvent":         reflect.ValueOf((*ToolOutputEvent)(nil)),
 			"ToolResultEvent":         reflect.ValueOf((*ToolResultEvent)(nil)),
 			"ToolResultResult":        reflect.ValueOf((*ToolResultResult)(nil)),
 			"InputEvent":              reflect.ValueOf((*InputEvent)(nil)),

internal/extensions/test_api.go

@@ -30,6 +30,12 @@ func NewTestAPI(ext *LoadedExtension) API {
 				return nil
 			})
 		},
+		onToolOutput: func(h func(ToolOutputEvent, Context)) {
+			reg(ToolOutput, func(e Event, c Context) Result {
+				h(e.(ToolOutputEvent), c)
+				return nil
+			})
+		},
 		onToolResult: func(h func(ToolResultEvent, Context) *ToolResultResult) {
 			reg(ToolResult, func(e Event, c Context) Result {
 				r := h(e.(ToolResultEvent), c)

internal/models/providers.go

@@ -37,15 +37,38 @@ func resolveModelAlias(provider, modelName string) string {
 	registry := GetGlobalRegistry()
 
 	aliasMap := map[string]string{
-		"claude-opus-latest":     "claude-opus-4-20250514",
-		"claude-sonnet-latest":   "claude-sonnet-4-5-20250929",
-		"claude-4-opus-latest":   "claude-opus-4-20250514",
-		"claude-4-sonnet-latest": "claude-sonnet-4-5-20250929",
-
+		// Anthropic aliases
+		"claude-opus-latest":       "claude-opus-4-6",
+		"claude-sonnet-latest":     "claude-sonnet-4-6",
+		"claude-haiku-latest":      "claude-haiku-4-5",
+		"claude-4-opus-latest":     "claude-opus-4-6",
+		"claude-4-sonnet-latest":   "claude-sonnet-4-6",
+		"claude-4-haiku-latest":    "claude-haiku-4-5",
 		"claude-3-5-haiku-latest":  "claude-3-5-haiku-20241022",
 		"claude-3-5-sonnet-latest": "claude-3-5-sonnet-20241022",
 		"claude-3-7-sonnet-latest": "claude-3-7-sonnet-20250219",
 		"claude-3-opus-latest":     "claude-3-opus-20240229",
+
+		// OpenAI aliases
+		"gpt-5-latest":      "gpt-5.4",
+		"gpt-5-chat-latest": "gpt-5.4",
+		"gpt-4-latest":      "gpt-4o",
+		"gpt-4":             "gpt-4o",
+		"gpt-3.5":           "gpt-3.5-turbo",
+		"gpt-3.5-latest":    "gpt-3.5-turbo",
+		"o1-latest":         "o1",
+		"o3-latest":         "o3",
+		"o4-latest":         "o4-mini",
+		"codex-latest":      "codex-mini-latest",
+
+		// Google Gemini aliases
+		"gemini-pro-latest": "gemini-2.5-pro",
+		"gemini-flash":      "gemini-2.5-flash",
+		"gemini-pro":        "gemini-2.5-pro",
+		"gemini-2-flash":    "gemini-2.0-flash",
+		"gemini-2-pro":      "gemini-2.5-pro",
+		"gemini-1.5-flash":  "gemini-1.5-flash",
+		"gemini-1.5-pro":    "gemini-1.5-pro",
 	}
 
 	if resolved, exists := aliasMap[modelName]; exists {
@@ -180,8 +203,8 @@ func CreateProvider(ctx context.Context, config *ProviderConfig) (*ProviderResul
 		return nil, err
 	}
 
-	// Resolve model aliases (for OAuth compatibility)
-	if provider == "anthropic" || provider == "google-vertex-anthropic" {
+	// Resolve model aliases to full model names
+	if provider == "anthropic" || provider == "google-vertex-anthropic" || provider == "openai" || provider == "google" {
 		modelName = resolveModelAlias(provider, modelName)
 	}
 

internal/prompts/frontmatter.go

@@ -0,0 +1,25 @@
+package prompts
+
+import (
+	"fmt"
+
+	"gopkg.in/yaml.v3"
+)
+
+// frontmatterSep is the YAML frontmatter delimiter.
+const frontmatterSep = "---"
+
+// Frontmatter represents the YAML frontmatter in a prompt template file.
+type Frontmatter struct {
+	// Description summarises what this template provides.
+	Description string `yaml:"description"`
+}
+
+// ParseFrontmatter parses YAML frontmatter content into a Frontmatter struct.
+func ParseFrontmatter(content string) (*Frontmatter, error) {
+	var fm Frontmatter
+	if err := yaml.Unmarshal([]byte(content), &fm); err != nil {
+		return nil, fmt.Errorf("parsing frontmatter: %w", err)
+	}
+	return &fm, nil
+}

internal/prompts/loader.go

@@ -0,0 +1,217 @@
+package prompts
+
+import (
+	"fmt"
+	"os"
+	"path/filepath"
+	"strings"
+
+	"github.com/charmbracelet/log"
+)
+
+// LoadOptions configures how templates are discovered and loaded.
+type LoadOptions struct {
+	// Cwd is the current working directory for project-local discovery.
+	// If empty, the current working directory is used.
+	Cwd string
+	// HomeDir is the user's home directory. If empty, os.UserHomeDir() is used.
+	HomeDir string
+	// ExtraPaths are additional explicit paths to search for templates.
+	ExtraPaths []string
+	// ConfigPaths are paths from configuration files to search.
+	ConfigPaths []string
+	// IncludeDefaults determines whether to include built-in default templates.
+	IncludeDefaults bool
+}
+
+// Diagnostic reports a template collision or loading issue.
+type Diagnostic struct {
+	// Name is the template name that had a collision.
+	Name string
+	// KeptPath is the path of the template that was kept (higher precedence).
+	KeptPath string
+	// DroppedPath is the path of the template that was dropped.
+	DroppedPath string
+	// Reason explains why the collision occurred.
+	Reason string
+}
+
+// LoadAll discovers and loads all prompt templates from standard locations
+// and any extra paths. Templates are loaded in order of precedence (lowest
+// to highest), with later templates overriding earlier ones of the same name.
+//
+// Discovery paths searched in order:
+//  1. Default templates (if IncludeDefaults)
+//  2. ~/.kit/prompts/ (global user templates)
+//  3. .kit/prompts/ (project-local templates)
+//  4. ConfigPaths (from configuration)
+//  5. ExtraPaths (explicit paths, highest precedence)
+func LoadAll(opts LoadOptions) ([]*PromptTemplate, []Diagnostic, error) {
+	if opts.Cwd == "" {
+		opts.Cwd, _ = os.Getwd()
+	}
+
+	if opts.HomeDir == "" {
+		home, err := os.UserHomeDir()
+		if err != nil {
+			return nil, nil, fmt.Errorf("getting home directory: %w", err)
+		}
+		opts.HomeDir = home
+	}
+
+	var all []*PromptTemplate
+	var diagnostics []Diagnostic
+	seen := make(map[string]*PromptTemplate) // name -> template
+
+	// Helper to add templates with deduplication tracking
+	addTemplates := func(templates []*PromptTemplate, source string) {
+		for _, tpl := range templates {
+			if existing, ok := seen[tpl.Name]; ok {
+				// Collision: report diagnostic, keep existing (lower precedence wins)
+				diagnostics = append(diagnostics, Diagnostic{
+					Name:        tpl.Name,
+					KeptPath:    existing.FilePath,
+					DroppedPath: tpl.FilePath,
+					Reason:      fmt.Sprintf("template from %s overridden by %s", source, existing.Source),
+				})
+				log.Debug("template collision",
+					"name", tpl.Name,
+					"dropped", tpl.FilePath,
+					"kept", existing.FilePath)
+			} else {
+				tpl.Source = source
+				seen[tpl.Name] = tpl
+				all = append(all, tpl)
+			}
+		}
+	}
+
+	// 1. Default templates (lowest precedence)
+	if opts.IncludeDefaults {
+		defaults := loadDefaultTemplates()
+		addTemplates(defaults, "default")
+	}
+
+	// 2. Global user templates: ~/.kit/prompts/
+	globalDir := filepath.Join(opts.HomeDir, ".kit", "prompts")
+	if templates, err := LoadFromDir(globalDir); err == nil {
+		addTemplates(templates, "global")
+	}
+
+	// 3. Project-local templates: .kit/prompts/
+	localDir := filepath.Join(opts.Cwd, ".kit", "prompts")
+	if templates, err := LoadFromDir(localDir); err == nil {
+		addTemplates(templates, "local")
+	}
+
+	// 4. Config paths
+	for _, path := range opts.ConfigPaths {
+		info, err := os.Stat(path)
+		if err != nil {
+			continue
+		}
+		if info.IsDir() {
+			if templates, err := LoadFromDir(path); err == nil {
+				addTemplates(templates, "config")
+			}
+		} else if strings.HasSuffix(path, ".md") {
+			if tpl, err := ParseTemplate(path); err == nil {
+				addTemplates([]*PromptTemplate{tpl}, "config")
+			}
+		}
+	}
+
+	// 5. Extra paths (highest precedence)
+	for _, path := range opts.ExtraPaths {
+		info, err := os.Stat(path)
+		if err != nil {
+			continue
+		}
+		if info.IsDir() {
+			if templates, err := LoadFromDir(path); err == nil {
+				addTemplates(templates, "explicit")
+			}
+		} else if strings.HasSuffix(path, ".md") {
+			if tpl, err := ParseTemplate(path); err == nil {
+				addTemplates([]*PromptTemplate{tpl}, "explicit")
+			}
+		}
+	}
+
+	return all, diagnostics, nil
+}
+
+// LoadFromDir scans a directory for .md files and loads them as templates.
+// It looks for *.md files directly in the directory.
+// Files that fail to parse are logged and skipped.
+func LoadFromDir(dir string) ([]*PromptTemplate, error) {
+	info, err := os.Stat(dir)
+	if err != nil || !info.IsDir() {
+		return nil, nil // directory doesn't exist — not an error
+	}
+
+	entries, err := os.ReadDir(dir)
+	if err != nil {
+		return nil, fmt.Errorf("reading prompts directory %s: %w", dir, err)
+	}
+
+	var templates []*PromptTemplate
+	var errs []string
+
+	for _, entry := range entries {
+		if entry.IsDir() {
+			continue
+		}
+
+		name := entry.Name()
+		if !strings.HasSuffix(name, ".md") {
+			continue
+		}
+
+		full := filepath.Join(dir, name)
+		tpl, err := ParseTemplate(full)
+		if err != nil {
+			errs = append(errs, err.Error())
+			continue
+		}
+		templates = append(templates, tpl)
+	}
+
+	if len(errs) > 0 {
+		return templates, fmt.Errorf("some templates failed to load: %s", strings.Join(errs, "; "))
+	}
+	return templates, nil
+}
+
+// Deduplicate removes duplicate templates by name, keeping the first occurrence.
+// It returns the deduplicated list and diagnostics for any collisions.
+// This is a standalone function for when you need to deduplicate an existing list.
+func Deduplicate(templates []*PromptTemplate) ([]*PromptTemplate, []Diagnostic) {
+	seen := make(map[string]*PromptTemplate)
+	var result []*PromptTemplate
+	var diagnostics []Diagnostic
+
+	for _, tpl := range templates {
+		if existing, ok := seen[tpl.Name]; ok {
+			diagnostics = append(diagnostics, Diagnostic{
+				Name:        tpl.Name,
+				KeptPath:    existing.FilePath,
+				DroppedPath: tpl.FilePath,
+				Reason:      "duplicate template name (first-match-wins)",
+			})
+		} else {
+			seen[tpl.Name] = tpl
+			result = append(result, tpl)
+		}
+	}
+
+	return result, diagnostics
+}
+
+// loadDefaultTemplates returns the built-in default templates.
+// These are embedded templates that ship with Kit.
+func loadDefaultTemplates() []*PromptTemplate {
+	// Default templates can be added here as needed
+	// For now, return an empty slice - users can define their own templates
+	return nil
+}

internal/prompts/loader_test.go

@@ -0,0 +1,126 @@
+package prompts
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+)
+
+func TestLoadAll_Integration(t *testing.T) {
+	// Create a temp directory for testing
+	tempDir := t.TempDir()
+
+	// Create the .kit/prompts subdirectory structure
+	promptsDir := filepath.Join(tempDir, ".kit", "prompts")
+	if err := os.MkdirAll(promptsDir, 0755); err != nil {
+		t.Fatalf("Failed to create prompts dir: %v", err)
+	}
+
+	// Create a test template file
+	templateContent := `---
+description: Test template for integration
+---
+Review $1 with focus on $2`
+
+	testFile := filepath.Join(promptsDir, "test.md")
+	if err := os.WriteFile(testFile, []byte(templateContent), 0644); err != nil {
+		t.Fatalf("Failed to create test file: %v", err)
+	}
+
+	// Test loading from the temp directory
+	tpls, diags, err := LoadAll(LoadOptions{
+		HomeDir:         tempDir,
+		IncludeDefaults: false, // Skip default locations for this test
+	})
+
+	if err != nil {
+		t.Fatalf("LoadAll failed: %v", err)
+	}
+
+	if len(diags) > 0 {
+		t.Logf("Got %d diagnostics", len(diags))
+	}
+
+	if len(tpls) != 1 {
+		t.Fatalf("Expected 1 template, got %d", len(tpls))
+	}
+
+	tpl := tpls[0]
+	if tpl.Name != "test" {
+		t.Errorf("Expected name 'test', got '%s'", tpl.Name)
+	}
+
+	if tpl.Description != "Test template for integration" {
+		t.Errorf("Expected description 'Test template for integration', got '%s'", tpl.Description)
+	}
+
+	// Test expansion
+	expanded := tpl.Expand("code security")
+	expected := "Review code with focus on security"
+	if expanded != expected {
+		t.Errorf("Expected '%s', got '%s'", expected, expanded)
+	}
+}
+
+func TestParseTemplate_WithFrontmatter(t *testing.T) {
+	// Create a temp file with frontmatter
+	tempDir := t.TempDir()
+	templateContent := `---
+description: A test template
+---
+Create a $1 component with $2 features`
+
+	testFile := filepath.Join(tempDir, "component.md")
+	if err := os.WriteFile(testFile, []byte(templateContent), 0644); err != nil {
+		t.Fatalf("Failed to create test file: %v", err)
+	}
+
+	tpl, err := ParseTemplate(testFile)
+	if err != nil {
+		t.Fatalf("ParseTemplate failed: %v", err)
+	}
+
+	if tpl.Name != "component" {
+		t.Errorf("Expected name 'component', got '%s'", tpl.Name)
+	}
+
+	if tpl.Description != "A test template" {
+		t.Errorf("Expected description 'A test template', got '%s'", tpl.Description)
+	}
+
+	expectedContent := "Create a $1 component with $2 features"
+	if tpl.Content != expectedContent {
+		t.Errorf("Expected content '%s', got '%s'", expectedContent, tpl.Content)
+	}
+}
+
+func TestParseTemplate_WithoutFrontmatter(t *testing.T) {
+	// Create a temp file without frontmatter
+	tempDir := t.TempDir()
+	templateContent := `Simple template without frontmatter
+Supports $1 and $2 placeholders`
+
+	testFile := filepath.Join(tempDir, "simple.md")
+	if err := os.WriteFile(testFile, []byte(templateContent), 0644); err != nil {
+		t.Fatalf("Failed to create test file: %v", err)
+	}
+
+	tpl, err := ParseTemplate(testFile)
+	if err != nil {
+		t.Fatalf("ParseTemplate failed: %v", err)
+	}
+
+	if tpl.Name != "simple" {
+		t.Errorf("Expected name 'simple', got '%s'", tpl.Name)
+	}
+
+	// Description should be empty since there's no frontmatter
+	if tpl.Description != "" {
+		t.Errorf("Expected empty description, got '%s'", tpl.Description)
+	}
+
+	// Content should include everything
+	if tpl.Content != templateContent {
+		t.Errorf("Content mismatch\nExpected:\n%s\nGot:\n%s", templateContent, tpl.Content)
+	}
+}

internal/prompts/template.go

@@ -0,0 +1,279 @@
+package prompts
+
+import (
+	"fmt"
+	"os"
+	"path/filepath"
+	"regexp"
+	"strconv"
+	"strings"
+)
+
+// PromptTemplate is a named prompt template with shell-style argument placeholders.
+// It supports Pi-style $1, $2, $@, $ARGUMENTS, ${@:N}, ${@:N:L} syntax.
+type PromptTemplate struct {
+	// Name is the human-readable identifier for this template.
+	Name string
+	// Description summarises what this template provides.
+	Description string
+	// Content is the raw template text with placeholders.
+	Content string
+	// Source indicates where the template was loaded from (e.g., "default", "user").
+	Source string
+	// FilePath is the absolute filesystem path the template was loaded from.
+	FilePath string
+}
+
+// ParseTemplate reads a template from a file. The template name is derived
+// from the filename (without extension). If the file contains YAML frontmatter,
+// the description is extracted from it.
+func ParseTemplate(path string) (*PromptTemplate, error) {
+	data, err := os.ReadFile(path)
+	if err != nil {
+		return nil, fmt.Errorf("reading template %s: %w", path, err)
+	}
+
+	abs, err := filepath.Abs(path)
+	if err != nil {
+		abs = path
+	}
+
+	content := string(data)
+	tpl := &PromptTemplate{
+		FilePath: abs,
+		Content:  content,
+	}
+
+	// Parse frontmatter if present
+	if strings.HasPrefix(strings.TrimSpace(content), frontmatterSep) {
+		trimmed := strings.TrimSpace(content)
+		rest := trimmed[len(frontmatterSep):]
+		frontmatter, body, found := strings.Cut(rest, "\n"+frontmatterSep)
+		if found {
+			body = strings.TrimPrefix(body, "\n")
+			fm, err := ParseFrontmatter(frontmatter)
+			if err == nil {
+				tpl.Description = fm.Description
+			}
+			tpl.Content = strings.TrimSpace(body)
+		}
+	}
+
+	// Derive name from filename
+	base := filepath.Base(path)
+	ext := filepath.Ext(base)
+	tpl.Name = strings.TrimSuffix(base, ext)
+
+	return tpl, nil
+}
+
+// ParseCommandArgs splits a command line into arguments respecting quotes.
+// It handles single quotes, double quotes, and backslash escaping.
+func ParseCommandArgs(input string) []string {
+	var args []string
+	var current strings.Builder
+	inSingleQuote := false
+	inDoubleQuote := false
+	escaped := false
+
+	for i, r := range input {
+		if escaped {
+			current.WriteRune(r)
+			escaped = false
+			continue
+		}
+
+		if r == '\\' && !inSingleQuote {
+			// Backslash escapes next char, but not in single quotes
+			escaped = true
+			continue
+		}
+
+		if r == '\'' && !inDoubleQuote {
+			inSingleQuote = !inSingleQuote
+			continue
+		}
+
+		if r == '"' && !inSingleQuote {
+			inDoubleQuote = !inDoubleQuote
+			continue
+		}
+
+		if r == ' ' && !inSingleQuote && !inDoubleQuote {
+			if current.Len() > 0 {
+				args = append(args, current.String())
+				current.Reset()
+			}
+			continue
+		}
+
+		current.WriteRune(r)
+		_ = i // silence unused warning when we need position later
+	}
+
+	if current.Len() > 0 {
+		args = append(args, current.String())
+	}
+
+	return args
+}
+
+// argPlaceholder matches shell-style argument placeholders:
+//   - $1, $2, etc. - positional arguments
+//   - $@ - all arguments
+//   - $ARGUMENTS - all arguments (alias for $@)
+//   - ${@:N} - arguments from N onwards
+//   - ${@:N:L} - L arguments starting from N
+var argPlaceholder = regexp.MustCompile(`\$\{(\d+)\}|\$\{(\d+):(\d+)\}|\$\{ARGUMENTS\}|\$\{@(:\d+)?(:\d+)?\}|\$(\d+)|\$@|\$ARGUMENTS`)
+
+// SubstituteArgs replaces argument placeholders in content with values from args.
+// Supported placeholders:
+//   - $N, ${N} - the Nth argument (1-indexed)
+//   - $@, $ARGUMENTS, ${ARGUMENTS} - all arguments joined with spaces
+//   - ${@:N} - arguments from index N onwards (0-indexed)
+//   - ${@:N:L} - L arguments starting from index N (0-indexed)
+func SubstituteArgs(content string, args []string) string {
+	return argPlaceholder.ReplaceAllStringFunc(content, func(match string) string {
+		// Check for ${N} or ${N:M} format
+		if strings.HasPrefix(match, "${") && strings.Contains(match, "}") {
+			inner := match[2 : len(match)-1] // Remove ${ and }
+
+			// Check for ${ARGUMENTS}
+			if inner == "ARGUMENTS" {
+				return strings.Join(args, " ")
+			}
+
+			// Check for ${@...} format
+			if strings.HasPrefix(inner, "@") {
+				return expandAtArgs(inner, args)
+			}
+
+			// Check for ${N:M} format (positional with length)
+			if colonIdx := strings.Index(inner, ":"); colonIdx > 0 {
+				startStr := inner[:colonIdx]
+				rest := inner[colonIdx+1:]
+
+				start, err := strconv.Atoi(startStr)
+				if err != nil || start < 1 {
+					return match
+				}
+
+				// Check if there's a second colon for length ${N:M:L}
+				lengthStr, _, ok := strings.Cut(rest, ":")
+				if ok {
+					length, err := strconv.Atoi(lengthStr)
+					if err != nil || length < 0 {
+						return match
+					}
+					return joinArgsRange(args, start-1, length)
+				}
+
+				// Single colon ${N:M} - M is length
+				length, err := strconv.Atoi(rest)
+				if err != nil || length < 0 {
+					return match
+				}
+				return joinArgsRange(args, start-1, length)
+			}
+
+			// Simple ${N} format
+			n, err := strconv.Atoi(inner)
+			if err != nil || n < 1 {
+				return match
+			}
+			if n <= len(args) {
+				return args[n-1]
+			}
+			return ""
+		}
+
+		// Check for $N format (without braces)
+		if strings.HasPrefix(match, "$") && !strings.HasPrefix(match, "${") {
+			suffix := match[1:]
+
+			// $@ or $ARGUMENTS
+			if suffix == "@" || suffix == "ARGUMENTS" {
+				return strings.Join(args, " ")
+			}
+
+			// $N
+			n, err := strconv.Atoi(suffix)
+			if err != nil || n < 1 {
+				return match
+			}
+			if n <= len(args) {
+				return args[n-1]
+			}
+			return ""
+		}
+
+		return match
+	})
+}
+
+// expandAtArgs handles ${@...} patterns (1-indexed like bash)
+func expandAtArgs(inner string, args []string) string {
+	// Remove the @ prefix
+	rest := inner[1:]
+
+	if rest == "" {
+		// ${@} - all arguments
+		return strings.Join(args, " ")
+	}
+
+	// Must start with :
+	if !strings.HasPrefix(rest, ":") {
+		return "${" + inner + "}"
+	}
+	rest = rest[1:]
+
+	// Parse start index
+	startStr, lengthStr, hasLength := strings.Cut(rest, ":")
+
+	start, err := strconv.Atoi(startStr)
+	if err != nil || start < 0 {
+		return "${" + inner + "}"
+	}
+
+	// Convert from 1-indexed to 0-indexed (bash convention)
+	// Treat 0 as 1 (bash convention: args start at 1)
+	if start > 0 {
+		start--
+	}
+
+	if hasLength {
+		length, err := strconv.Atoi(lengthStr)
+		if err != nil || length < 0 {
+			return "${" + inner + "}"
+		}
+		return joinArgsRange(args, start, length)
+	}
+
+	// ${@:N} - from N to end
+	if start >= len(args) {
+		return ""
+	}
+	return strings.Join(args[start:], " ")
+}
+
+// joinArgsRange joins args from start index, taking up to length elements
+func joinArgsRange(args []string, start, length int) string {
+	if start >= len(args) || length <= 0 {
+		return ""
+	}
+	end := start + length
+	end = min(end, len(args))
+	return strings.Join(args[start:end], " ")
+}
+
+// Expand substitutes arguments into the template content and returns the result.
+// It first parses args from the input string, then substitutes them into the template.
+func (t *PromptTemplate) Expand(argsInput string) string {
+	args := ParseCommandArgs(argsInput)
+	return SubstituteArgs(t.Content, args)
+}
+
+// ExpandWithArgs substitutes the provided arguments into the template content.
+func (t *PromptTemplate) ExpandWithArgs(args []string) string {
+	return SubstituteArgs(t.Content, args)
+}

internal/prompts/template_test.go

@@ -0,0 +1,215 @@
+package prompts
+
+import (
+	"testing"
+)
+
+func TestParseCommandArgs(t *testing.T) {
+	tests := []struct {
+		input    string
+		expected []string
+	}{
+		{"", []string{}},
+		{"hello", []string{"hello"}},
+		{"hello world", []string{"hello", "world"}},
+		{`"hello world"`, []string{"hello world"}},
+		{`'hello world'`, []string{"hello world"}},
+		{`hello "world foo" bar`, []string{"hello", "world foo", "bar"}},
+		{`hello 'world foo' bar`, []string{"hello", "world foo", "bar"}},
+		{`hello \"world\"`, []string{"hello", `"world"`}},
+		{`hello \\world`, []string{"hello", `\world`}},
+		{`  hello   world  `, []string{"hello", "world"}},
+		{`Button "onClick handler" "disabled support"`, []string{"Button", "onClick handler", "disabled support"}},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.input, func(t *testing.T) {
+			got := ParseCommandArgs(tt.input)
+			if len(got) != len(tt.expected) {
+				t.Errorf("ParseCommandArgs(%q) = %v, want %v", tt.input, got, tt.expected)
+				return
+			}
+			for i := range got {
+				if got[i] != tt.expected[i] {
+					t.Errorf("ParseCommandArgs(%q)[%d] = %q, want %q", tt.input, i, got[i], tt.expected[i])
+				}
+			}
+		})
+	}
+}
+
+func TestSubstituteArgs(t *testing.T) {
+	tests := []struct {
+		name     string
+		content  string
+		args     []string
+		expected string
+	}{
+		{
+			name:     "no placeholders",
+			content:  "Hello world",
+			args:     []string{},
+			expected: "Hello world",
+		},
+		{
+			name:     "positional $1",
+			content:  "Hello $1",
+			args:     []string{"world"},
+			expected: "Hello world",
+		},
+		{
+			name:     "positional $1 $2",
+			content:  "$1 and $2",
+			args:     []string{"first", "second"},
+			expected: "first and second",
+		},
+		{
+			name:     "missing arg",
+			content:  "Hello $1 and $2",
+			args:     []string{"world"},
+			expected: "Hello world and ",
+		},
+		{
+			name:     "$@ wildcard",
+			content:  "Args: $@",
+			args:     []string{"a", "b", "c"},
+			expected: "Args: a b c",
+		},
+		{
+			name:     "$ARGUMENTS wildcard",
+			content:  "Args: $ARGUMENTS",
+			args:     []string{"a", "b", "c"},
+			expected: "Args: a b c",
+		},
+		{
+			name:     "${@} all args",
+			content:  "Args: ${@}",
+			args:     []string{"a", "b", "c"},
+			expected: "Args: a b c",
+		},
+		{
+			name:     "${@:2} slice from index 2",
+			content:  "Rest: ${@:2}",
+			args:     []string{"a", "b", "c", "d"},
+			expected: "Rest: b c d",
+		},
+		{
+			name:     "${@:1:2} slice with length",
+			content:  "First two: ${@:1:2}",
+			args:     []string{"a", "b", "c", "d"},
+			expected: "First two: a b",
+		},
+		{
+			name:     "${@:0} from start",
+			content:  "All: ${@:0}",
+			args:     []string{"a", "b", "c"},
+			expected: "All: a b c",
+		},
+		{
+			name:     "${@:3:1} single arg",
+			content:  "Third: ${@:3:1}",
+			args:     []string{"a", "b", "c", "d"},
+			expected: "Third: c",
+		},
+		{
+			name:     "combined placeholders",
+			content:  "Create $1 with features: $ARGUMENTS",
+			args:     []string{"Button", "onClick", "disabled"},
+			expected: "Create Button with features: Button onClick disabled",
+		},
+		{
+			name:     "slice beyond bounds",
+			content:  "${@:10}",
+			args:     []string{"a", "b"},
+			expected: "",
+		},
+		{
+			name:     "empty args with wildcard",
+			content:  "Args: $@",
+			args:     []string{},
+			expected: "Args: ",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			got := SubstituteArgs(tt.content, tt.args)
+			if got != tt.expected {
+				t.Errorf("SubstituteArgs(%q, %v) = %q, want %q", tt.content, tt.args, got, tt.expected)
+			}
+		})
+	}
+}
+
+func TestParseFrontmatter(t *testing.T) {
+	tests := []struct {
+		name     string
+		content  string
+		wantDesc string
+		wantErr  bool
+	}{
+		{
+			name:     "simple description",
+			content:  "description: Review code\n",
+			wantDesc: "Review code",
+		},
+		{
+			name:     "empty",
+			content:  "",
+			wantDesc: "",
+		},
+		{
+			name:     "invalid yaml",
+			content:  "description: [unclosed",
+			wantDesc: "",
+			wantErr:  true,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			fm, err := ParseFrontmatter(tt.content)
+			if (err != nil) != tt.wantErr {
+				t.Errorf("ParseFrontmatter() error = %v, wantErr %v", err, tt.wantErr)
+				return
+			}
+			if err != nil {
+				return
+			}
+			if fm.Description != tt.wantDesc {
+				t.Errorf("ParseFrontmatter() Description = %q, want %q", fm.Description, tt.wantDesc)
+			}
+		})
+	}
+}
+
+func TestPromptTemplateExpand(t *testing.T) {
+	tpl := &PromptTemplate{
+		Name:        "component",
+		Description: "Create a component",
+		Content:     "Create a React component named $1 with features: $ARGUMENTS",
+	}
+
+	tests := []struct {
+		input    string
+		expected string
+	}{
+		{
+			input:    "Button",
+			expected: "Create a React component named Button with features: Button",
+		},
+		{
+			input:    `Button "onClick handler"`,
+			expected: "Create a React component named Button with features: Button onClick handler",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.input, func(t *testing.T) {
+			got := tpl.Expand(tt.input)
+			if got != tt.expected {
+				t.Errorf("Expand(%q) = %q, want %q", tt.input, got, tt.expected)
+			}
+		})
+	}
+}

internal/session/entry.go

@@ -23,6 +23,7 @@ const (
 	EntryTypeLabel         EntryType = "label"
 	EntryTypeSessionInfo   EntryType = "session_info"
 	EntryTypeExtensionData EntryType = "extension_data"
+	EntryTypeCompaction    EntryType = "compaction"
 )
 
 // CurrentVersion is the session format version for JSONL tree sessions.
@@ -102,6 +103,20 @@ type ExtensionDataEntry struct {
 	Data    string `json:"data"`     // Extension-defined data (JSON or plain text)
 }
 
+// CompactionEntry records an LLM-generated summary of older messages.
+// Instead of deleting old messages, the tree manager skips entries before
+// FirstKeptEntryID when building the LLM context, preserving full history.
+type CompactionEntry struct {
+	Entry
+	Summary          string   `json:"summary"`
+	FirstKeptEntryID string   `json:"first_kept_entry_id"`
+	TokensBefore     int      `json:"tokens_before"`
+	TokensAfter      int      `json:"tokens_after"`
+	MessagesRemoved  int      `json:"messages_removed"`
+	ReadFiles        []string `json:"read_files,omitempty"`
+	ModifiedFiles    []string `json:"modified_files,omitempty"`
+}
+
 // GenerateEntryID creates a unique entry identifier (16 hex chars).
 func GenerateEntryID() string {
 	bytes := make([]byte, 8)
@@ -188,6 +203,20 @@ func NewExtensionDataEntry(parentID, extType, data string) *ExtensionDataEntry {
 	}
 }
 
+// NewCompactionEntry creates a CompactionEntry.
+func NewCompactionEntry(parentID, summary, firstKeptEntryID string, tokensBefore, tokensAfter, messagesRemoved int, readFiles, modifiedFiles []string) *CompactionEntry {
+	return &CompactionEntry{
+		Entry:            NewEntry(EntryTypeCompaction, parentID),
+		Summary:          summary,
+		FirstKeptEntryID: firstKeptEntryID,
+		TokensBefore:     tokensBefore,
+		TokensAfter:      tokensAfter,
+		MessagesRemoved:  messagesRemoved,
+		ReadFiles:        readFiles,
+		ModifiedFiles:    modifiedFiles,
+	}
+}
+
 // --- JSONL marshaling helpers ---
 
 // MarshalEntry serializes any entry to a JSON line (no trailing newline).
@@ -259,6 +288,13 @@ func UnmarshalEntry(data []byte) (any, error) {
 		}
 		return &e, nil
 
+	case EntryTypeCompaction:
+		var e CompactionEntry
+		if err := json.Unmarshal(data, &e); err != nil {
+			return nil, fmt.Errorf("failed to unmarshal compaction entry: %w", err)
+		}
+		return &e, nil
+
 	default:
 		return nil, fmt.Errorf("unknown entry type: %q", env.Type)
 	}

internal/session/tree_manager.go

@@ -4,6 +4,7 @@ import (
 	"bufio"
 	"encoding/json"
 	"fmt"
+	"io"
 	"os"
 	"path/filepath"
 	"strings"
@@ -128,10 +129,34 @@ func OpenTreeSession(path string) (*TreeManager, error) {
 		filePath:   path,
 	}
 
-	scanner := bufio.NewScanner(strings.NewReader(string(data)))
+	reader := bufio.NewReader(strings.NewReader(string(data)))
 	lineNum := 0
-	for scanner.Scan() {
-		line := scanner.Text()
+	for {
+		line, err := reader.ReadString('\n')
+		if err != nil {
+			if err == io.EOF {
+				// Process the last line if it's not empty
+				if strings.TrimSpace(line) != "" {
+					lineNum++
+					entry, err := UnmarshalEntry([]byte(line))
+					if err != nil {
+						return nil, fmt.Errorf("line %d: %w", lineNum, err)
+					}
+					if lineNum == 1 {
+						h, ok := entry.(*SessionHeader)
+						if !ok {
+							return nil, fmt.Errorf("first line must be a session header, got %T", entry)
+						}
+						tm.header = *h
+					} else {
+						tm.addEntryToIndex(entry)
+					}
+				}
+				break
+			}
+			return nil, fmt.Errorf("failed to read session file: %w", err)
+		}
+
 		if strings.TrimSpace(line) == "" {
 			continue
 		}
@@ -153,9 +178,6 @@ func OpenTreeSession(path string) (*TreeManager, error) {
 
 		tm.addEntryToIndex(entry)
 	}
-	if err := scanner.Err(); err != nil {
-		return nil, fmt.Errorf("failed to scan session file: %w", err)
-	}
 
 	// Set leaf to the last entry.
 	if len(tm.entries) > 0 {
@@ -298,6 +320,22 @@ func (tm *TreeManager) AppendExtensionData(extType, data string) (string, error)
 	return entry.ID, nil
 }
 
+// AppendCompaction adds a compaction entry to the tree. The entry records
+// the summary and the ID of the first entry that should be preserved in the
+// LLM context. Messages before that entry are replaced by the summary.
+func (tm *TreeManager) AppendCompaction(summary, firstKeptEntryID string, tokensBefore, tokensAfter, messagesRemoved int, readFiles, modifiedFiles []string) (string, error) {
+	tm.mu.Lock()
+	defer tm.mu.Unlock()
+
+	entry := NewCompactionEntry(tm.leafID, summary, firstKeptEntryID, tokensBefore, tokensAfter, messagesRemoved, readFiles, modifiedFiles)
+	if err := tm.appendAndPersist(entry); err != nil {
+		return "", err
+	}
+
+	tm.leafID = entry.ID
+	return entry.ID, nil
+}
+
 // GetExtensionData returns all extension data entries matching the given type,
 // walking the current branch from root to leaf. If extType is empty, all
 // extension data entries on the branch are returned.
@@ -441,8 +479,9 @@ func (tm *TreeManager) GetTree() []*TreeNode {
 // --- Context building ---
 
 // BuildContext walks from the current leaf to the root and returns the
-// conversation messages suitable for sending to the LLM. Branch summaries
-// are converted to user messages to provide context from abandoned branches.
+// conversation messages suitable for sending to the LLM. Compaction entries
+// cause older messages to be replaced by the summary. Branch summaries are
+// converted to user messages to provide context from abandoned branches.
 // Also returns the latest model/provider settings encountered on the path.
 func (tm *TreeManager) BuildContext() (messages []fantasy.Message, provider string, modelID string) {
 	tm.mu.RLock()
@@ -455,7 +494,41 @@ func (tm *TreeManager) BuildContext() (messages []fantasy.Message, provider stri
 	// Walk from leaf to root collecting entries.
 	branch := tm.getBranchLocked(tm.leafID)
 
+	// Find the last compaction entry on this branch — it determines
+	// which older messages are replaced by the summary.
+	var lastCompaction *CompactionEntry
+	for i := len(branch) - 1; i >= 0; i-- {
+		if c, ok := branch[i].(*CompactionEntry); ok {
+			lastCompaction = c
+			break
+		}
+	}
+
+	// If there is a compaction, inject the summary first.
+	if lastCompaction != nil {
+		messages = append(messages, fantasy.Message{
+			Role: fantasy.MessageRoleSystem,
+			Content: []fantasy.MessagePart{
+				fantasy.TextPart{
+					Text: fmt.Sprintf("[Conversation summary — earlier messages were compacted]\n\n%s", lastCompaction.Summary),
+				},
+			},
+		})
+	}
+
+	// Determine whether to skip entries (everything before firstKeptEntryID).
+	skipping := lastCompaction != nil
 	for _, entry := range branch {
+		// Once we reach the first kept entry, stop skipping.
+		if skipping {
+			entryID := tm.entryID(entry)
+			if entryID == lastCompaction.FirstKeptEntryID {
+				skipping = false
+			} else {
+				continue
+			}
+		}
+
 		switch e := entry.(type) {
 		case *MessageEntry:
 			msg, err := e.ToMessage()
@@ -481,6 +554,10 @@ func (tm *TreeManager) BuildContext() (messages []fantasy.Message, provider stri
 		case *ModelChangeEntry:
 			provider = e.Provider
 			modelID = e.ModelID
+
+		case *CompactionEntry:
+			// Already handled above (the last one on the branch).
+			continue
 		}
 	}
 
@@ -563,6 +640,96 @@ func (tm *TreeManager) Close() error {
 	return nil
 }
 
+// GetContextEntryIDs returns the entry IDs corresponding to the fantasy
+// messages returned by BuildContext, in the same order. Each entry ID maps
+// to the session entry that produced the fantasy message at the same index.
+// This is used by compaction to map a cut point index back to an entry ID.
+//
+// Note: A single MessageEntry produces at most one fantasy message. Branch
+// summary entries also produce one message each. The returned slice has the
+// same length as the messages slice from BuildContext (excluding the
+// compaction summary system message, which has no entry ID — it gets the
+// empty string "").
+func (tm *TreeManager) GetContextEntryIDs() []string {
+	tm.mu.RLock()
+	defer tm.mu.RUnlock()
+
+	if tm.leafID == "" {
+		return nil
+	}
+
+	branch := tm.getBranchLocked(tm.leafID)
+
+	// Find the last compaction entry for skip logic.
+	var lastCompaction *CompactionEntry
+	for i := len(branch) - 1; i >= 0; i-- {
+		if c, ok := branch[i].(*CompactionEntry); ok {
+			lastCompaction = c
+			break
+		}
+	}
+
+	var ids []string
+
+	// If there's a compaction summary injected, it has no entry ID.
+	if lastCompaction != nil {
+		ids = append(ids, "") // placeholder for the summary system message
+	}
+
+	skipping := lastCompaction != nil
+	for _, entry := range branch {
+		if skipping {
+			entryID := tm.entryID(entry)
+			if entryID == lastCompaction.FirstKeptEntryID {
+				skipping = false
+			} else {
+				continue
+			}
+		}
+
+		switch e := entry.(type) {
+		case *MessageEntry:
+			msg, err := e.ToMessage()
+			if err != nil {
+				continue
+			}
+			msgs := msg.ToFantasyMessages()
+			for range msgs {
+				ids = append(ids, e.ID)
+			}
+
+		case *BranchSummaryEntry:
+			if e.Summary != "" {
+				ids = append(ids, e.ID)
+			}
+
+		case *CompactionEntry:
+			continue
+		}
+	}
+
+	return ids
+}
+
+// GetLastCompaction returns the most recent CompactionEntry on the current
+// branch, or nil if none exists. Used to carry forward file tracking.
+func (tm *TreeManager) GetLastCompaction() *CompactionEntry {
+	tm.mu.RLock()
+	defer tm.mu.RUnlock()
+
+	if tm.leafID == "" {
+		return nil
+	}
+
+	branch := tm.getBranchLocked(tm.leafID)
+	for i := len(branch) - 1; i >= 0; i-- {
+		if c, ok := branch[i].(*CompactionEntry); ok {
+			return c
+		}
+	}
+	return nil
+}
+
 // --- Legacy bridge ---
 
 // AddFantasyMessages appends multiple fantasy messages as entries. This is
@@ -641,6 +808,8 @@ func (tm *TreeManager) entryID(entry any) string {
 		return e.ID
 	case *ExtensionDataEntry:
 		return e.ID
+	case *CompactionEntry:
+		return e.ID
 	default:
 		return ""
 	}
@@ -661,6 +830,8 @@ func (tm *TreeManager) entryParentID(entry any) string {
 		return e.ParentID
 	case *ExtensionDataEntry:
 		return e.ParentID
+	case *CompactionEntry:
+		return e.ParentID
 	default:
 		return ""
 	}

internal/ui/commands.go

@@ -141,6 +141,27 @@ var SlashCommands = []SlashCommand{
 		Description: "Set a display name for this session",
 		Category:    "Navigation",
 	},
+	{
+		Name:        "/resume",
+		Description: "Open session picker to switch sessions",
+		Category:    "Navigation",
+		Aliases:     []string{"/r"},
+	},
+	{
+		Name:        "/export",
+		Description: "Export session (JSONL by default, or /export path.jsonl)",
+		Category:    "System",
+	},
+	{
+		Name:        "/share",
+		Description: "Share session via GitHub Gist (requires gh CLI)",
+		Category:    "System",
+	},
+	{
+		Name:        "/import",
+		Description: "Import a session from a JSONL file (/import path.jsonl)",
+		Category:    "System",
+	},
 	{
 		Name:        "/session",
 		Description: "Show session info and statistics",

internal/ui/compact_renderer.go

@@ -82,9 +82,20 @@ func (r *CompactRenderer) RenderUserMessage(content string, timestamp time.Time)
 }
 
 // RenderAssistantMessage renders an AI assistant's response in compact format with
-// a distinctive symbol (<) and the model name as label. Empty content is displayed
-// as "(no output)". Returns a UIMessage with formatted content and metadata.
+// a distinctive symbol (<) and the model name as label. Empty content is ignored
+// and returns an empty message. Returns a UIMessage with formatted content and metadata.
 func (r *CompactRenderer) RenderAssistantMessage(content string, timestamp time.Time, modelName string) UIMessage {
+	// Ignore empty responses - don't render anything
+	compactContent := r.formatUserAssistantContent(content)
+	if compactContent == "" {
+		return UIMessage{
+			Type:      AssistantMessage,
+			Content:   "",
+			Height:    0,
+			Timestamp: timestamp,
+		}
+	}
+
 	theme := getTheme()
 	symbol := lipgloss.NewStyle().Foreground(theme.Primary).Render("<")
 
@@ -94,12 +105,6 @@ func (r *CompactRenderer) RenderAssistantMessage(content string, timestamp time.
 	}
 	label := lipgloss.NewStyle().Foreground(theme.Primary).Bold(true).Render(modelName)
 
-	// Format content for assistant messages (preserve formatting, no truncation)
-	compactContent := r.formatUserAssistantContent(content)
-	if compactContent == "" {
-		compactContent = lipgloss.NewStyle().Foreground(theme.Muted).Italic(true).Render("(no output)")
-	}
-
 	// Handle multi-line content
 	lines := strings.Split(compactContent, "\n")
 	var formattedLines []string

internal/ui/input.go

@@ -68,8 +68,26 @@ type InputComponent struct {
 	// pendingImages holds clipboard images attached to the next submission.
 	// Images are added via Ctrl+V and cleared on submit or Ctrl+U.
 	pendingImages []ImageAttachment
+
+	// history stores previously submitted prompts (most recent last).
+	// Limited to maxHistory entries; duplicates of the previous entry are
+	// skipped. Empty strings are never stored.
+	history []string
+	// historyIndex is the current position when browsing history.
+	// When not browsing, historyIndex == len(history).
+	historyIndex int
+	// savedInput holds the user's in-progress text before they started
+	// browsing history, so it can be restored when they press down past
+	// the end of history.
+	savedInput string
+	// browsingHistory is true when the user is navigating history with
+	// up/down arrows. Set to false when they type a character or submit.
+	browsingHistory bool
 }
 
+// maxHistory is the maximum number of prompt entries kept in history.
+const maxHistory = 100
+
 // clipboardImageMsg is the result of an async clipboard image read.
 type clipboardImageMsg struct {
 	image *ImageAttachment
@@ -138,6 +156,7 @@ func (s *InputComponent) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 	if s.submitNext {
 		s.submitNext = false
 		value := s.textarea.Value()
+		s.pushHistory(value)
 		s.textarea.SetValue("")
 		s.textarea.CursorEnd()
 		s.showPopup = false
@@ -166,10 +185,47 @@ func (s *InputComponent) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 			switch msg.String() {
 			case "ctrl+d", "enter":
 				value := s.textarea.Value()
+				s.pushHistory(value)
 				s.textarea.SetValue("")
 				s.textarea.CursorEnd()
 				s.lastValue = ""
 				return s, s.handleSubmit(value)
+			case "up":
+				// Navigate prompt history backward (older entries).
+				if len(s.history) > 0 {
+					if !s.browsingHistory {
+						// Start browsing — save current input.
+						s.savedInput = s.textarea.Value()
+						s.browsingHistory = true
+						s.historyIndex = len(s.history)
+					}
+					if s.historyIndex > 0 {
+						s.historyIndex--
+						s.textarea.SetValue(s.history[s.historyIndex])
+						s.textarea.CursorEnd()
+						s.lastValue = s.textarea.Value()
+					}
+					return s, nil
+				}
+			case "down":
+				// Navigate prompt history forward (newer entries).
+				if s.browsingHistory {
+					if s.historyIndex < len(s.history)-1 {
+						s.historyIndex++
+						s.textarea.SetValue(s.history[s.historyIndex])
+						s.textarea.CursorEnd()
+						s.lastValue = s.textarea.Value()
+					} else {
+						// Past the end — restore saved input.
+						s.historyIndex = len(s.history)
+						s.browsingHistory = false
+						s.textarea.SetValue(s.savedInput)
+						s.textarea.CursorEnd()
+						s.lastValue = s.textarea.Value()
+						s.savedInput = ""
+					}
+					return s, nil
+				}
 			case "ctrl+v":
 				// Try to read an image from the clipboard asynchronously.
 				return s, readClipboardImageCmd()
@@ -250,6 +306,11 @@ func (s *InputComponent) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 		value := s.textarea.Value()
 		if value != s.lastValue {
 			s.lastValue = value
+			// User typed something — exit history browsing mode.
+			if s.browsingHistory {
+				s.browsingHistory = false
+				s.savedInput = ""
+			}
 			lines := strings.Split(value, "\n")
 			line := lines[len(lines)-1] // current line (last line for multi-line)
 
@@ -372,6 +433,34 @@ func (s *InputComponent) handleSubmit(value string) tea.Cmd {
 	}
 }
 
+// pushHistory adds a prompt to the history ring buffer. Empty strings and
+// consecutive duplicates of the last entry are skipped. When the buffer
+// exceeds maxHistory, the oldest entry is dropped.
+func (s *InputComponent) pushHistory(value string) {
+	trimmed := strings.TrimSpace(value)
+	if trimmed == "" {
+		return
+	}
+	// Skip consecutive duplicates.
+	if len(s.history) > 0 && s.history[len(s.history)-1] == trimmed {
+		s.resetHistoryBrowsing()
+		return
+	}
+	s.history = append(s.history, trimmed)
+	if len(s.history) > maxHistory {
+		s.history = s.history[len(s.history)-maxHistory:]
+	}
+	s.resetHistoryBrowsing()
+}
+
+// resetHistoryBrowsing resets the history browsing state so the index
+// points past the end (ready for new input).
+func (s *InputComponent) resetHistoryBrowsing() {
+	s.historyIndex = len(s.history)
+	s.browsingHistory = false
+	s.savedInput = ""
+}
+
 // View implements tea.Model. Renders the title, textarea, autocomplete popup
 // (if visible), and help text.
 func (s *InputComponent) View() tea.View {

internal/ui/messages.go

@@ -213,22 +213,21 @@ func (r *MessageRenderer) RenderUserMessage(content string, timestamp time.Time)
 
 // RenderAssistantMessage renders an AI assistant's response with left-aligned formatting,
 // including the model name, timestamp, and markdown-rendered content. Empty responses
-// are displayed with a special "Finished without output" message. The message features
-// a colored left border for visual distinction.
+// are ignored and return an empty message. The message features a colored left border
+// for visual distinction.
 func (r *MessageRenderer) RenderAssistantMessage(content string, timestamp time.Time, modelName string) UIMessage {
-	theme := getTheme()
-
-	var messageContent string
+	// Ignore empty responses - don't render anything
 	if strings.TrimSpace(content) == "" {
-		messageContent = lipgloss.NewStyle().
-			Italic(true).
-			Foreground(theme.Muted).
-			Align(lipgloss.Center).
-			Render("Finished without output")
-	} else {
-		messageContent = r.renderMarkdown(content, r.width-8)
+		return UIMessage{
+			Type:      AssistantMessage,
+			Content:   "",
+			Height:    0,
+			Timestamp: timestamp,
+		}
 	}
 
+	theme := getTheme()
+	messageContent := r.renderMarkdown(content, r.width-8)
 	fullContent := strings.TrimSuffix(messageContent, "\n")
 
 	// Left border with Primary (Mauve) color for assistant messages.

internal/ui/model.go

@@ -7,6 +7,7 @@ import (
 	"os"
 	"os/exec"
 	"strings"
+	"sync"
 	"time"
 
 	tea "charm.land/bubbletea/v2"
@@ -15,7 +16,9 @@ import (
 
 	"github.com/mark3labs/kit/internal/app"
 	"github.com/mark3labs/kit/internal/core"
+	"github.com/mark3labs/kit/internal/message"
 	"github.com/mark3labs/kit/internal/models"
+	"github.com/mark3labs/kit/internal/prompts"
 	"github.com/mark3labs/kit/internal/session"
 )
 
@@ -44,6 +47,9 @@ const (
 
 	// stateModelSelector means the /model selector overlay is active.
 	stateModelSelector
+
+	// stateSessionSelector means the /resume session picker is active.
+	stateSessionSelector
 )
 
 // AppController is the interface the parent TUI model uses to interact with the
@@ -245,6 +251,11 @@ type AppModelOptions struct {
 	// appear in autocomplete, /help, and are dispatched when submitted.
 	ExtensionCommands []ExtensionCommand
 
+	// PromptTemplates are user-defined prompt templates loaded from ~/.kit/prompts/,
+	// .kit/prompts/, or explicit --prompt-template paths. They appear in autocomplete
+	// and are expanded when submitted (e.g., /review → full prompt text).
+	PromptTemplates []*prompts.PromptTemplate
+
 	// ContextPaths lists absolute paths of loaded context files (e.g.
 	// AGENTS.md). Displayed in the [Context] startup section.
 	ContextPaths []string
@@ -330,6 +341,16 @@ type AppModelOptions struct {
 	// May be nil if extensions are not loaded.
 	EmitModelChange func(newModel, previousModel, source string)
 
+	// SwitchSession opens a session by JSONL file path, replacing the
+	// active tree session and reloading messages. Called when the user
+	// picks a session from /resume. May be nil if session switching is
+	// not supported.
+	SwitchSession func(path string) error
+
+	// ShowSessionPicker, when true, opens the session picker immediately
+	// on startup (used by --resume flag).
+	ShowSessionPicker bool
+
 	// ThinkingLevel is the initial thinking level (e.g. "off", "medium").
 	ThinkingLevel string
 	// IsReasoningModel is true when the current model supports reasoning.
@@ -430,6 +451,10 @@ type AppModel struct {
 	// handleExtensionCommand when submitted.
 	extensionCommands []ExtensionCommand
 
+	// promptTemplates are user-defined prompt templates for expansion.
+	// They appear in autocomplete and are expanded when submitted.
+	promptTemplates []*prompts.PromptTemplate
+
 	// treeSelector is the tree navigation overlay, active in stateTreeSelector.
 	treeSelector *TreeSelectorComponent
 
@@ -497,6 +522,13 @@ type AppModel struct {
 	// modelSelector is the model selection overlay, active in stateModelSelector.
 	modelSelector *ModelSelectorComponent
 
+	// sessionSelector is the session picker overlay, active in stateSessionSelector.
+	sessionSelector *SessionSelectorComponent
+
+	// switchSession opens a session by JSONL path, replacing the active session.
+	// Wired from cmd/root.go.
+	switchSession func(path string) error
+
 	// prompt holds the state of an active interactive prompt overlay. Nil
 	// when no prompt is active. Managed by updatePromptState().
 	prompt *promptOverlay
@@ -529,6 +561,16 @@ type AppModel struct {
 	// width and height track the terminal dimensions.
 	width  int
 	height int
+
+	// streamingBashOutput holds the current streaming bash output lines.
+	// Lines are accumulated as they arrive and displayed in the stream region.
+	streamingBashOutput []string
+	// streamingBashStderr holds stderr lines separately (rendered differently).
+	streamingBashStderr []string
+	// streamingBashMaxLines caps how many lines to accumulate to prevent memory issues.
+	streamingBashMaxLines int
+	// streamingMu protects the streaming bash output fields from concurrent access.
+	streamingMu sync.RWMutex
 }
 
 // --------------------------------------------------------------------------
@@ -614,6 +656,7 @@ func NewAppModel(appCtrl AppController, opts AppModelOptions) *AppModel {
 
 	// Store extension commands for dispatch.
 	m.extensionCommands = opts.ExtensionCommands
+	m.promptTemplates = opts.PromptTemplates
 	m.getWidgets = opts.GetWidgets
 	m.getHeader = opts.GetHeader
 	m.getFooter = opts.GetFooter
@@ -630,6 +673,7 @@ func NewAppModel(appCtrl AppController, opts AppModelOptions) *AppModel {
 	m.thinkingVisible = true // default to showing thinking blocks
 	m.isReasoningModel = opts.IsReasoningModel
 	m.setThinkingLevel = opts.SetThinkingLevel
+	m.switchSession = opts.SwitchSession
 
 	// Store context/skills metadata and tool counts for startup display.
 	m.contextPaths = opts.ContextPaths
@@ -637,6 +681,9 @@ func NewAppModel(appCtrl AppController, opts AppModelOptions) *AppModel {
 	m.mcpToolCount = opts.MCPToolCount
 	m.extensionToolCount = opts.ExtensionToolCount
 
+	// Initialize streaming bash output buffer.
+	m.streamingBashMaxLines = 50 // cap to prevent memory issues
+
 	// Wire up child components now that we have the concrete implementations.
 	m.input = NewInputComponent(width, "Enter your prompt (Type /help for commands, Ctrl+C to quit)", appCtrl)
 
@@ -657,9 +704,26 @@ func NewAppModel(appCtrl AppController, opts AppModelOptions) *AppModel {
 		}
 	}
 
+	// Merge prompt templates into the InputComponent's autocomplete source.
+	if ic, ok := m.input.(*InputComponent); ok && len(opts.PromptTemplates) > 0 {
+		for _, tpl := range opts.PromptTemplates {
+			ic.commands = append(ic.commands, SlashCommand{
+				Name:        "/" + tpl.Name,
+				Description: tpl.Description,
+				Category:    "Prompts",
+			})
+		}
+	}
+
 	m.stream = NewStreamComponent(opts.CompactMode, width, opts.ModelName)
 	m.stream.SetThinkingVisible(m.thinkingVisible)
 
+	// If --resume was passed, open the session picker immediately.
+	if opts.ShowSessionPicker {
+		m.sessionSelector = NewSessionSelector(opts.Cwd, width, height)
+		m.state = stateSessionSelector
+	}
+
 	// Propagate initial height distribution to children.
 	m.distributeHeight()
 
@@ -852,6 +916,8 @@ func (m *AppModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 					m.modelName = parts[1]
 				}
 				m.printSystemMessage(fmt.Sprintf("Switched to %s", msg.ModelString))
+				// Persist model selection for next launch.
+				go func() { _ = SaveModelPreference(msg.ModelString) }()
 				if m.emitModelChange != nil {
 					emit := m.emitModelChange
 					newModel := msg.ModelString
@@ -868,6 +934,34 @@ func (m *AppModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 		m.state = stateInput
 		return m, nil
 
+	// ── Session selector events ──────────────────────────────────────────────
+	case SessionSelectedMsg:
+		m.sessionSelector = nil
+		m.state = stateInput
+		if m.switchSession != nil {
+			if err := m.switchSession(msg.Path); err != nil {
+				m.printSystemMessage(fmt.Sprintf("Failed to switch session: %v", err))
+			} else {
+				m.renderSessionHistory()
+				m.printSystemMessage("Session loaded. Continue where you left off.")
+			}
+		} else {
+			m.printSystemMessage("Session switching not available.")
+		}
+		cmds = append(cmds, m.drainScrollback())
+		return m, tea.Batch(cmds...)
+
+	case SessionSelectorCancelledMsg:
+		m.sessionSelector = nil
+		m.state = stateInput
+		return m, nil
+
+	case SessionDeletedMsg:
+		// Session was deleted from picker — just show a message.
+		m.printSystemMessage(fmt.Sprintf("Deleted session: %s", msg.Name))
+		cmds = append(cmds, m.drainScrollback())
+		return m, tea.Batch(cmds...)
+
 	// ── Window resize ────────────────────────────────────────────────────────
 	case tea.WindowSizeMsg:
 		m.width = msg.Width
@@ -951,6 +1045,14 @@ func (m *AppModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 			return m, tea.Batch(cmds...)
 		}
 
+		// Route to session selector when active.
+		if m.state == stateSessionSelector && m.sessionSelector != nil {
+			updated, cmd := m.sessionSelector.Update(msg)
+			m.sessionSelector = updated.(*SessionSelectorComponent)
+			cmds = append(cmds, cmd)
+			return m, tea.Batch(cmds...)
+		}
+
 		switch msg.String() {
 		case "esc":
 			if m.state == stateWorking {
@@ -1065,6 +1167,24 @@ func (m *AppModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 					}
 					cmds = append(cmds, m.drainScrollback())
 					return m, tea.Batch(cmds...)
+				case "/name":
+					if cmd := m.handleNameCommand(strings.TrimSpace(args)); cmd != nil {
+						cmds = append(cmds, cmd)
+					}
+					cmds = append(cmds, m.drainScrollback())
+					return m, tea.Batch(cmds...)
+				case "/export":
+					if cmd := m.handleExportCommand(strings.TrimSpace(args)); cmd != nil {
+						cmds = append(cmds, cmd)
+					}
+					cmds = append(cmds, m.drainScrollback())
+					return m, tea.Batch(cmds...)
+				case "/import":
+					if cmd := m.handleImportCommand(strings.TrimSpace(args)); cmd != nil {
+						cmds = append(cmds, cmd)
+					}
+					cmds = append(cmds, m.drainScrollback())
+					return m, tea.Batch(cmds...)
 				}
 			}
 		}
@@ -1076,6 +1196,12 @@ func (m *AppModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 			return m, tea.Batch(cmds...)
 		}
 
+		// Expand prompt templates. If the input matches a template name,
+		// substitute arguments and use the expanded content as the prompt.
+		if expanded, ok := m.expandPromptTemplate(msg.Text); ok {
+			msg.Text = expanded
+		}
+
 		// Regular prompt — forward to the app layer.
 		// Preprocess @file references: expand them into XML-wrapped file
 		// content before sending to the agent. The display text (shown in
@@ -1200,12 +1326,35 @@ func (m *AppModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 	case app.ToolResultEvent:
 		// Buffer tool result for scrollback.
 		m.printToolResult(msg)
+		// Clear streaming bash output since tool completed.
+		m.streamingMu.Lock()
+		m.streamingBashOutput = nil
+		m.streamingBashStderr = nil
+		m.streamingMu.Unlock()
 		// Start spinner again while waiting for the next LLM response.
 		if m.stream != nil {
 			_, cmd := m.stream.Update(app.SpinnerEvent{Show: true})
 			cmds = append(cmds, cmd)
 		}
 
+	case app.ToolOutputEvent:
+		// Accumulate streaming bash output for display.
+		m.streamingMu.Lock()
+		if msg.IsStderr {
+			m.streamingBashStderr = append(m.streamingBashStderr, msg.Chunk)
+			// Cap stderr lines to prevent memory issues.
+			if len(m.streamingBashStderr) > m.streamingBashMaxLines {
+				m.streamingBashStderr = m.streamingBashStderr[len(m.streamingBashStderr)-m.streamingBashMaxLines:]
+			}
+		} else {
+			m.streamingBashOutput = append(m.streamingBashOutput, msg.Chunk)
+			// Cap stdout lines to prevent memory issues.
+			if len(m.streamingBashOutput) > m.streamingBashMaxLines {
+				m.streamingBashOutput = m.streamingBashOutput[len(m.streamingBashOutput)-m.streamingBashMaxLines:]
+			}
+		}
+		m.streamingMu.Unlock()
+
 	case app.ToolCallContentEvent:
 		// In streaming mode this text was already delivered via StreamChunkEvents
 		// and will be flushed before the next tool call. Ignore to avoid
@@ -1218,7 +1367,7 @@ func (m *AppModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 		// print or reset — flushStreamContent() handles it on the next step.
 		// In non-streaming mode (no stream content accumulated), print the text.
 		hasStreamContent := m.stream != nil && m.stream.GetRenderedContent() != ""
-		if !hasStreamContent && msg.Content != "" {
+		if !hasStreamContent && strings.TrimSpace(msg.Content) != "" {
 			m.printAssistantMessage(msg.Content)
 			if m.stream != nil {
 				m.stream.Reset()
@@ -1421,6 +1570,14 @@ func (m *AppModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 			cmds = append(cmds, m.performFork(msg.targetID, msg.isUser, msg.userText))
 		}
 
+	case shareResultMsg:
+		if msg.err != nil {
+			m.printSystemMessage(fmt.Sprintf("Share failed: %v", msg.err))
+		} else {
+			m.printSystemMessage(fmt.Sprintf("Session shared!\n\n  Viewer: %s\n  Gist:   %s", msg.viewerURL, msg.gistURL))
+		}
+		return m, m.drainScrollback()
+
 	case app.ExtensionPrintEvent:
 		// Extension output — route through styled renderers when a level is set.
 		switch msg.Level {
@@ -1467,6 +1624,11 @@ func (m *AppModel) View() tea.View {
 		return m.modelSelector.View()
 	}
 
+	// Session selector overlay replaces the normal layout.
+	if m.state == stateSessionSelector && m.sessionSelector != nil {
+		return m.sessionSelector.View()
+	}
+
 	// Overlay dialog replaces the normal layout.
 	if m.state == stateOverlay && m.overlay != nil {
 		return tea.NewView(m.overlay.Render())
@@ -1545,24 +1707,80 @@ func (m *AppModel) View() tea.View {
 
 // renderStream returns the stream region content.
 func (m *AppModel) renderStream() string {
-	if m.stream == nil {
+	theme := GetTheme()
+
+	var parts []string
+
+	// Stream component content (LLM streaming text, reasoning, spinner placeholder).
+	if m.stream != nil {
+		if content := m.stream.View().Content; content != "" {
+			parts = append(parts, content)
+		}
+	}
+
+	// Streaming bash output section (if any).
+	bashView := m.renderStreamingBashOutput(theme)
+	if bashView != "" {
+		parts = append(parts, bashView)
+	}
+
+	if len(parts) == 0 {
 		return ""
 	}
 
 	// Show canceling warning if set.
 	if m.canceling {
-		theme := GetTheme()
 		warning := lipgloss.NewStyle().
 			Foreground(theme.Warning).
 			Bold(true).
 			Render("  ⚠ Press ESC again to cancel")
-		return lipgloss.JoinVertical(lipgloss.Left,
-			m.stream.View().Content,
-			warning,
-		)
+		parts = append(parts, warning)
 	}
 
-	return m.stream.View().Content
+	return lipgloss.JoinVertical(lipgloss.Left, parts...)
+}
+
+// renderStreamingBashOutput renders accumulated streaming bash output (stdout + stderr)
+// below the LLM streaming text. Returns empty string if no bash output is present.
+func (m *AppModel) renderStreamingBashOutput(theme Theme) string {
+	m.streamingMu.RLock()
+	stdoutLines := make([]string, len(m.streamingBashOutput))
+	copy(stdoutLines, m.streamingBashOutput)
+	stderrLines := make([]string, len(m.streamingBashStderr))
+	copy(stderrLines, m.streamingBashStderr)
+	m.streamingMu.RUnlock()
+
+	if len(stdoutLines) == 0 && len(stderrLines) == 0 {
+		return ""
+	}
+
+	const lineIndent = "  "
+	width := m.width - 2 // Account for indent and padding
+
+	outputStyle := lipgloss.NewStyle().
+		Background(theme.CodeBg).
+		PaddingLeft(1)
+
+	stderrStyle := lipgloss.NewStyle().
+		Foreground(theme.Error).
+		Background(theme.CodeBg).
+		PaddingLeft(1)
+
+	var lines []string
+
+	// Render stdout lines.
+	for _, line := range stdoutLines {
+		styled := outputStyle.Width(width - len(lineIndent)).Render(line)
+		lines = append(lines, lineIndent+styled)
+	}
+
+	// Render stderr lines with error styling.
+	for _, line := range stderrLines {
+		styled := stderrStyle.Width(width - len(lineIndent)).Render(line)
+		lines = append(lines, lineIndent+styled)
+	}
+
+	return strings.Join(lines, "\n")
 }
 
 // renderStatusBar renders a persistent single-line status bar below the input.
@@ -1678,6 +1896,9 @@ func (m *AppModel) cycleThinkingLevel() {
 			_ = m.setThinkingLevel(next)
 		}()
 	}
+
+	// Persist thinking level for next launch.
+	go func() { _ = SaveThinkingLevelPreference(next) }()
 }
 
 // renderSeparator renders the separator line with an optional queue count badge.
@@ -1823,7 +2044,7 @@ func (m *AppModel) printUserMessage(text string) {
 
 // printAssistantMessage renders an assistant message into the scrollback buffer.
 func (m *AppModel) printAssistantMessage(text string) {
-	if text != "" {
+	if strings.TrimSpace(text) != "" {
 		m.appendScrollback(m.renderer.RenderAssistantMessage(text, time.Now(), m.modelName).Content)
 	}
 }
@@ -1888,7 +2109,15 @@ func (m *AppModel) handleSlashCommand(sc *SlashCommand) tea.Cmd {
 	case "/new":
 		return m.handleNewCommand()
 	case "/name":
-		return m.handleNameCommand()
+		return m.handleNameCommand("")
+	case "/resume":
+		return m.handleResumeCommand()
+	case "/export":
+		return m.handleExportCommand("")
+	case "/share":
+		return m.handleShareCommand()
+	case "/import":
+		return m.handleImportCommand("")
 	case "/session":
 		return m.handleSessionInfoCommand()
 
@@ -1978,6 +2207,33 @@ func (m *AppModel) handleExtensionCommand(text string) tea.Cmd {
 	return func() tea.Msg { return nil }
 }
 
+// expandPromptTemplate checks if the submitted text matches a prompt template
+// and returns the expanded content with arguments substituted.
+// Returns (expanded, true) if a template was found and expanded, (text, false) otherwise.
+func (m *AppModel) expandPromptTemplate(text string) (string, bool) {
+	if len(m.promptTemplates) == 0 {
+		return text, false
+	}
+
+	// Only consider inputs that look like slash commands.
+	if !strings.HasPrefix(text, "/") {
+		return text, false
+	}
+
+	// Split: "/templatename arg1 arg2" → name="/templatename", args="arg1 arg2"
+	name, args, _ := strings.Cut(text, " ")
+	name = strings.TrimPrefix(name, "/")
+
+	// Find matching template
+	for _, tpl := range m.promptTemplates {
+		if tpl.Name == name {
+			return tpl.Expand(args), true
+		}
+	}
+
+	return text, false
+}
+
 // printHelpMessage renders the help text listing all available slash commands.
 func (m *AppModel) printHelpMessage() {
 	help := "## Available Commands\n\n" +
@@ -1990,10 +2246,14 @@ func (m *AppModel) printHelpMessage() {
 		"**Navigation:**\n" +
 		"- `/tree`: Navigate session tree (switch branches)\n" +
 		"- `/fork`: Branch from an earlier message\n" +
-		"- `/new`: Start a new branch (preserves history)\n\n" +
+		"- `/new`: Start a new branch (preserves history)\n" +
+		"- `/resume`: Open session picker to switch sessions\n" +
+		"- `/name <name>`: Set a display name for this session\n\n" +
 		"**System:**\n" +
 		"- `/compact [instructions]`: Summarise older messages to free context space\n" +
 		"- `/clear`: Clear message history\n" +
+		"- `/export [path]`: Export session as JSONL\n" +
+		"- `/import <path.jsonl>`: Import session from JSONL file\n" +
 		"- `/reset-usage`: Reset usage statistics\n" +
 		"- `/quit`: Exit the application\n\n"
 
@@ -2415,6 +2675,9 @@ func (m *AppModel) handleModelCommand(args string) tea.Cmd {
 		go emit(newModel, prev, "user")
 	}
 
+	// Persist model selection for next launch.
+	go func() { _ = SaveModelPreference(args) }()
+
 	m.printSystemMessage(fmt.Sprintf("Switched to %s", args))
 	return nil
 }
@@ -2504,6 +2767,8 @@ func (m *AppModel) handleThinkingCommand(args string) tea.Cmd {
 			_ = m.setThinkingLevel(string(level))
 		}()
 	}
+	// Persist thinking level for next launch.
+	go func() { _ = SaveThinkingLevelPreference(string(level)) }()
 	m.printSystemMessage(fmt.Sprintf("Thinking level set to: %s — %s", level, models.ThinkingLevelDescription(level)))
 	return nil
 }
@@ -2621,24 +2886,300 @@ func (m *AppModel) performFork(targetID string, isUser bool, userText string) te
 }
 
 // handleNameCommand sets a display name for the current session.
-func (m *AppModel) handleNameCommand() tea.Cmd {
+// Usage: /name <new name> — sets the session name.
+//
+//	/name             — shows the current name.
+func (m *AppModel) handleNameCommand(args string) tea.Cmd {
 	ts := m.appCtrl.GetTreeSession()
 	if ts == nil {
 		m.printSystemMessage("No tree session active.")
 		return nil
 	}
-	// For now, prompt user to provide name via input. We print instructions
-	// and the next non-command input starting with "name:" will be captured.
-	// TODO: inline input dialog.
-	currentName := ts.GetSessionName()
-	if currentName != "" {
-		m.printSystemMessage(fmt.Sprintf("Current session name: %q\nTo rename, type: `/name <new name>` (not yet implemented — use the session file directly).", currentName))
+
+	if args == "" {
+		// No argument — show current name.
+		currentName := ts.GetSessionName()
+		if currentName != "" {
+			m.printSystemMessage(fmt.Sprintf("Session name: %q\nTo rename: `/name <new name>`", currentName))
+		} else {
+			m.printSystemMessage("Session has no name. Set one with: `/name <new name>`")
+		}
 		return nil
 	}
-	m.printSystemMessage("To name this session, use: `/name <new name>` (not yet implemented — use the session file directly).")
+
+	// Set the session name.
+	if _, err := ts.AppendSessionInfo(args); err != nil {
+		m.printSystemMessage(fmt.Sprintf("Failed to set session name: %v", err))
+		return nil
+	}
+	m.printSystemMessage(fmt.Sprintf("Session named %q", args))
 	return nil
 }
 
+// handleExportCommand exports the current session to a file.
+// Usage: /export          — copies the JSONL file to cwd with a descriptive name.
+//
+//	/export path.jsonl — copies to the specified path.
+func (m *AppModel) handleExportCommand(args string) tea.Cmd {
+	ts := m.appCtrl.GetTreeSession()
+	if ts == nil {
+		m.printSystemMessage("No tree session active.")
+		return nil
+	}
+
+	srcPath := ts.GetFilePath()
+	if srcPath == "" {
+		m.printSystemMessage("Session is in-memory (not persisted). Nothing to export.")
+		return nil
+	}
+
+	// Determine destination path.
+	dstPath := args
+	if dstPath == "" {
+		// Generate a name based on session name or ID.
+		name := ts.GetSessionName()
+		if name == "" {
+			name = ts.GetSessionID()[:12]
+		}
+		// Sanitize for filename.
+		name = strings.Map(func(r rune) rune {
+			if r == '/' || r == '\\' || r == ':' || r == ' ' {
+				return '_'
+			}
+			return r
+		}, name)
+		dstPath = fmt.Sprintf("session_%s.jsonl", name)
+	}
+
+	// Copy the file.
+	data, err := os.ReadFile(srcPath)
+	if err != nil {
+		m.printSystemMessage(fmt.Sprintf("Failed to read session file: %v", err))
+		return nil
+	}
+
+	if err := os.WriteFile(dstPath, data, 0644); err != nil {
+		m.printSystemMessage(fmt.Sprintf("Failed to write export file: %v", err))
+		return nil
+	}
+
+	m.printSystemMessage(fmt.Sprintf("Session exported to: %s (%d bytes)", dstPath, len(data)))
+	return nil
+}
+
+// handleShareCommand uploads the current session as a GitHub Gist and prints
+// a shareable viewer URL. Requires the GitHub CLI (gh) to be installed and
+// authenticated.
+func (m *AppModel) handleShareCommand() tea.Cmd {
+	ts := m.appCtrl.GetTreeSession()
+	if ts == nil {
+		m.printSystemMessage("No tree session active.")
+		return nil
+	}
+
+	srcPath := ts.GetFilePath()
+	if srcPath == "" {
+		m.printSystemMessage("Session is in-memory (not persisted). Nothing to share.")
+		return nil
+	}
+
+	// Check that gh CLI is available.
+	if _, err := exec.LookPath("gh"); err != nil {
+		m.printSystemMessage("GitHub CLI (gh) is not installed. Install it from https://cli.github.com/")
+		return nil
+	}
+
+	// Check that gh is authenticated.
+	authCheck := exec.Command("gh", "auth", "status")
+	if err := authCheck.Run(); err != nil {
+		m.printSystemMessage("GitHub CLI is not logged in. Run 'gh auth login' first.")
+		return nil
+	}
+
+	// Copy session to a temp file with a clean name.
+	data, err := os.ReadFile(srcPath)
+	if err != nil {
+		m.printSystemMessage(fmt.Sprintf("Failed to read session file: %v", err))
+		return nil
+	}
+
+	name := ts.GetSessionName()
+	if name == "" {
+		name = "session"
+	}
+	// Sanitize for filename.
+	name = strings.Map(func(r rune) rune {
+		if r == '/' || r == '\\' || r == ':' || r == ' ' {
+			return '_'
+		}
+		return r
+	}, name)
+
+	tmpFile, err := os.CreateTemp("", fmt.Sprintf("kit-%s-*.jsonl", name))
+	if err != nil {
+		m.printSystemMessage(fmt.Sprintf("Failed to create temp file: %v", err))
+		return nil
+	}
+	tmpPath := tmpFile.Name()
+
+	if _, err := tmpFile.Write(data); err != nil {
+		_ = tmpFile.Close()
+		_ = os.Remove(tmpPath)
+		m.printSystemMessage(fmt.Sprintf("Failed to write temp file: %v", err))
+		return nil
+	}
+	_ = tmpFile.Close()
+
+	m.printSystemMessage("Uploading session to GitHub Gist...")
+
+	// Run gh gist create in background to avoid blocking the UI.
+	return func() tea.Msg {
+		defer func() { _ = os.Remove(tmpPath) }()
+
+		cmd := exec.Command("gh", "gist", "create", tmpPath, "--desc", "Kit session shared via /share")
+		output, err := cmd.Output()
+		if err != nil {
+			return shareResultMsg{err: fmt.Errorf("failed to create gist: %w", err)}
+		}
+
+		// gh outputs the gist URL like: https://gist.github.com/username/abc123def456
+		gistURL := strings.TrimSpace(string(output))
+
+		// Extract gist ID (last path segment).
+		parts := strings.Split(gistURL, "/")
+		gistID := parts[len(parts)-1]
+
+		viewerURL := fmt.Sprintf("https://go-kit.dev/session/#%s", gistID)
+		return shareResultMsg{gistURL: gistURL, viewerURL: viewerURL}
+	}
+}
+
+// handleImportCommand imports a session from a JSONL file.
+// Usage: /import path.jsonl
+func (m *AppModel) handleImportCommand(args string) tea.Cmd {
+	if args == "" {
+		m.printSystemMessage("Usage: `/import <path.jsonl>`")
+		return nil
+	}
+
+	if m.switchSession == nil {
+		m.printSystemMessage("Session switching is not available.")
+		return nil
+	}
+
+	// Verify file exists before attempting to switch.
+	if _, err := os.Stat(args); err != nil {
+		m.printSystemMessage(fmt.Sprintf("File not found: %s", args))
+		return nil
+	}
+
+	if err := m.switchSession(args); err != nil {
+		m.printSystemMessage(fmt.Sprintf("Failed to import session: %v", err))
+		return nil
+	}
+
+	m.renderSessionHistory()
+	m.printSystemMessage(fmt.Sprintf("Session imported from: %s", args))
+	return nil
+}
+
+// handleResumeCommand opens the session picker so the user can switch sessions.
+func (m *AppModel) handleResumeCommand() tea.Cmd {
+	if m.switchSession == nil {
+		m.printSystemMessage("Session switching is not available.")
+		return nil
+	}
+
+	m.sessionSelector = NewSessionSelector(m.cwd, m.width, m.height)
+	m.state = stateSessionSelector
+	return nil
+}
+
+// renderSessionHistory walks the current session branch and renders all
+// messages (user, assistant, tool calls/results) into the scrollback buffer.
+// This gives the user visual context of the conversation when resuming or
+// importing a session. Call this after switchSession succeeds.
+func (m *AppModel) renderSessionHistory() {
+	ts := m.appCtrl.GetTreeSession()
+	if ts == nil {
+		return
+	}
+
+	branch := ts.GetBranch("")
+	if len(branch) == 0 {
+		return
+	}
+
+	// First pass: build a map of tool call ID → {name, args} from assistant
+	// messages so we can pair them with tool results.
+	type toolCallInfo struct {
+		Name string
+		Args string
+	}
+	toolCallMap := make(map[string]toolCallInfo)
+	for _, entry := range branch {
+		me, ok := entry.(*session.MessageEntry)
+		if !ok {
+			continue
+		}
+		if me.Role != "assistant" {
+			continue
+		}
+		msg, err := me.ToMessage()
+		if err != nil {
+			continue
+		}
+		for _, tc := range msg.ToolCalls() {
+			toolCallMap[tc.ID] = toolCallInfo{Name: tc.Name, Args: tc.Input}
+		}
+	}
+
+	// Second pass: render each message in order.
+	for _, entry := range branch {
+		me, ok := entry.(*session.MessageEntry)
+		if !ok {
+			continue
+		}
+		msg, err := me.ToMessage()
+		if err != nil {
+			continue
+		}
+
+		switch msg.Role {
+		case message.RoleUser:
+			text := msg.Content()
+			if text != "" {
+				m.appendScrollback(m.renderer.RenderUserMessage(text, msg.CreatedAt).Content)
+			}
+
+		case message.RoleAssistant:
+			text := msg.Content()
+			if text != "" {
+				modelName := m.modelName
+				if msg.Model != "" {
+					modelName = msg.Model
+				}
+				m.appendScrollback(m.renderer.RenderAssistantMessage(text, msg.CreatedAt, modelName).Content)
+			}
+			// Tool calls from assistant messages are rendered when we
+			// encounter their corresponding tool results below.
+
+		case message.RoleTool:
+			for _, tr := range msg.ToolResults() {
+				toolName := tr.Name
+				toolArgs := ""
+				if info, ok := toolCallMap[tr.ToolCallID]; ok {
+					if toolName == "" {
+						toolName = info.Name
+					}
+					toolArgs = info.Args
+				}
+				m.appendScrollback(m.renderer.RenderToolMessage(toolName, toolArgs, tr.Content, tr.IsError).Content)
+			}
+		}
+	}
+}
+
 // handleSessionInfoCommand shows session statistics.
 func (m *AppModel) handleSessionInfoCommand() tea.Cmd {
 	ts := m.appCtrl.GetTreeSession()
@@ -2689,6 +3230,13 @@ func cancelTimerCmd() tea.Cmd {
 // Interactive prompt support
 // --------------------------------------------------------------------------
 
+// shareResultMsg carries the result of an async gist upload.
+type shareResultMsg struct {
+	err       error
+	gistURL   string
+	viewerURL string
+}
+
 // extensionCmdResultMsg carries the result of an asynchronously executed
 // extension slash command. Extension commands run async (via tea.Cmd) so they
 // can safely call blocking operations like ctx.PromptSelect().
@@ -2874,6 +3422,14 @@ func (m *AppModel) executeShellCommand(msg shellCommandMsg) tea.Cmd {
 			cmd.Dir = cwd
 		}
 
+		// Ensure SHELL is set to bash so child processes (e.g. tmux) use bash
+		// rather than the user's login shell (which may be nushell, fish, etc.).
+		bashPath, _ := exec.LookPath("bash")
+		if bashPath == "" {
+			bashPath = "/bin/bash"
+		}
+		cmd.Env = append(os.Environ(), "SHELL="+bashPath)
+
 		var stdout, stderr bytes.Buffer
 		cmd.Stdout = &stdout
 		cmd.Stderr = &stderr
@@ -2945,9 +3501,19 @@ func (m *AppModel) handleShellCommandResult(msg shellCommandResultMsg) tea.Cmd {
 	var displayHiddenCount int
 	if displayOutput != "" {
 		lines := strings.Split(displayOutput, "\n")
+		// Cap individual line length to prevent long lines from wrapping
+		// into excessive visual rows.
+		maxLineChars := max(m.width*3, 200)
+		for i, line := range lines {
+			if len(line) > maxLineChars {
+				lines[i] = line[:maxLineChars] + "…"
+			}
+		}
 		if len(lines) > maxShellDisplayLines {
 			displayHiddenCount = len(lines) - maxShellDisplayLines
 			displayOutput = strings.Join(lines[:maxShellDisplayLines], "\n")
+		} else {
+			displayOutput = strings.Join(lines, "\n")
 		}
 	}
 

internal/ui/model_test.go

@@ -112,15 +112,16 @@ func newTestAppModel(ctrl AppController) (*AppModel, *stubStreamComponent, *stub
 	stream := &stubStreamComponent{}
 	input := &stubInputComponent{}
 	m := &AppModel{
-		state:       stateInput,
-		appCtrl:     ctrl,
-		stream:      stream,
-		input:       input,
-		renderer:    newMessageRenderer(80, false),
-		compactMode: false,
-		modelName:   "test-model",
-		width:       80,
-		height:      24,
+		state:                 stateInput,
+		appCtrl:               ctrl,
+		stream:                stream,
+		input:                 input,
+		renderer:              newMessageRenderer(80, false),
+		compactMode:           false,
+		modelName:             "test-model",
+		width:                 80,
+		height:                24,
+		streamingBashMaxLines: 50, // Initialize buffer cap like NewAppModel does
 	}
 	return m, stream, input
 }
@@ -602,6 +603,82 @@ func TestToolResult_printsAndStartsSpinner(t *testing.T) {
 	}
 }
 
+// TestToolOutputEvent_accumulatesBashOutput verifies that ToolOutputEvent
+// accumulates stdout and stderr lines into the streaming bash output buffers.
+func TestToolOutputEvent_accumulatesBashOutput(t *testing.T) {
+	ctrl := &stubAppController{}
+	m, _, _ := newTestAppModel(ctrl)
+	m.state = stateWorking
+
+	// Send stdout chunk.
+	m = sendMsg(m, app.ToolOutputEvent{
+		ToolCallID: "call-1",
+		ToolName:   "bash",
+		Chunk:      "line one\n",
+		IsStderr:   false,
+	})
+
+	if len(m.streamingBashOutput) != 1 || m.streamingBashOutput[0] != "line one\n" {
+		t.Fatalf("expected streamingBashOutput=['line one\\n'], got %v", m.streamingBashOutput)
+	}
+	if len(m.streamingBashStderr) != 0 {
+		t.Fatalf("expected empty streamingBashStderr, got %v", m.streamingBashStderr)
+	}
+
+	// Send another stdout chunk.
+	m = sendMsg(m, app.ToolOutputEvent{
+		ToolCallID: "call-1",
+		ToolName:   "bash",
+		Chunk:      "line two\n",
+		IsStderr:   false,
+	})
+
+	if len(m.streamingBashOutput) != 2 {
+		t.Fatalf("expected 2 stdout lines, got %d", len(m.streamingBashOutput))
+	}
+
+	// Send stderr chunk.
+	m = sendMsg(m, app.ToolOutputEvent{
+		ToolCallID: "call-1",
+		ToolName:   "bash",
+		Chunk:      "error: something failed\n",
+		IsStderr:   true,
+	})
+
+	if len(m.streamingBashStderr) != 1 {
+		t.Fatalf("expected 1 stderr line, got %d", len(m.streamingBashStderr))
+	}
+	if m.streamingBashStderr[0] != "error: something failed\n" {
+		t.Fatalf("expected stderr 'error: something failed\\n', got %q", m.streamingBashStderr[0])
+	}
+}
+
+// TestToolResult_clearsStreamingBashOutput verifies that ToolResultEvent clears
+// the streaming bash output buffers since the final result will be printed.
+func TestToolResult_clearsStreamingBashOutput(t *testing.T) {
+	ctrl := &stubAppController{}
+	m, _, _ := newTestAppModel(ctrl)
+	m.state = stateWorking
+
+	// Accumulate some bash output.
+	m.streamingBashOutput = []string{"output line"}
+	m.streamingBashStderr = []string{"error line"}
+
+	_, _ = m.Update(app.ToolResultEvent{
+		ToolName: "bash",
+		ToolArgs: `{"cmd":"ls"}`,
+		Result:   "output line\nerror line\n",
+		IsError:  false,
+	})
+
+	if len(m.streamingBashOutput) != 0 {
+		t.Fatalf("expected streamingBashOutput cleared, got %v", m.streamingBashOutput)
+	}
+	if len(m.streamingBashStderr) != 0 {
+		t.Fatalf("expected streamingBashStderr cleared, got %v", m.streamingBashStderr)
+	}
+}
+
 // TestStepError_printCmd verifies that StepErrorEvent with a non-nil error
 // produces a non-nil cmd (the tea.Println call for the error message).
 func TestStepError_printCmd(t *testing.T) {

internal/ui/preferences.go

@@ -0,0 +1,129 @@
+package ui
+
+import (
+	"os"
+	"path/filepath"
+	"strings"
+
+	"gopkg.in/yaml.v3"
+)
+
+// preferences holds user-mutable runtime state that persists across sessions.
+// Stored at ~/.config/kit/preferences.yml, separate from the declarative
+// .kit.yml config so we never clobber user comments or formatting.
+type preferences struct {
+	Theme         string `yaml:"theme,omitempty"`
+	Model         string `yaml:"model,omitempty"`
+	ThinkingLevel string `yaml:"thinking_level,omitempty"`
+}
+
+// preferencesPath returns ~/.config/kit/preferences.yml.
+// Returns "" if the config directory cannot be determined.
+func preferencesPath() string {
+	cfgDir, err := os.UserConfigDir()
+	if err != nil {
+		return ""
+	}
+	return filepath.Join(cfgDir, "kit", "preferences.yml")
+}
+
+// loadPreferences reads and parses the preferences file.
+// Returns zero-value preferences if the file is missing or invalid.
+func loadPreferences() preferences {
+	path := preferencesPath()
+	if path == "" {
+		return preferences{}
+	}
+	data, err := os.ReadFile(path)
+	if err != nil {
+		return preferences{}
+	}
+	var prefs preferences
+	if err := yaml.Unmarshal(data, &prefs); err != nil {
+		return preferences{}
+	}
+	return prefs
+}
+
+// savePreferences atomically writes the preferences file, merging into any
+// existing content. The mutate function receives the current preferences and
+// should modify them in place.
+func savePreferences(mutate func(*preferences)) error {
+	path := preferencesPath()
+	if path == "" {
+		return nil // silently skip if config dir unavailable
+	}
+
+	// Load existing preferences to preserve other fields.
+	var prefs preferences
+	if data, err := os.ReadFile(path); err == nil {
+		_ = yaml.Unmarshal(data, &prefs)
+	}
+
+	mutate(&prefs)
+
+	data, err := yaml.Marshal(&prefs)
+	if err != nil {
+		return err
+	}
+
+	// Ensure parent directory exists.
+	if err := os.MkdirAll(filepath.Dir(path), 0o755); err != nil {
+		return err
+	}
+
+	// Atomic write: write to temp file, then rename.
+	tmp := path + ".tmp"
+	if err := os.WriteFile(tmp, data, 0o644); err != nil {
+		return err
+	}
+	return os.Rename(tmp, path)
+}
+
+// ── Theme preference ────────────────────────────────────────────────────────
+
+// LoadThemePreference reads the persisted theme name from preferences.yml.
+// Returns "" if no preference is saved or the file doesn't exist.
+func LoadThemePreference() string {
+	return strings.TrimSpace(loadPreferences().Theme)
+}
+
+// SaveThemePreference persists the theme name to ~/.config/kit/preferences.yml.
+// Preserves other preference fields. Uses atomic write (temp + rename) to
+// avoid corruption from concurrent Kit instances.
+func SaveThemePreference(name string) error {
+	return savePreferences(func(p *preferences) {
+		p.Theme = name
+	})
+}
+
+// ── Model preference ────────────────────────────────────────────────────────
+
+// LoadModelPreference reads the persisted model string (e.g.
+// "anthropic/claude-sonnet-4-5-20250929") from preferences.yml.
+// Returns "" if no preference is saved.
+func LoadModelPreference() string {
+	return strings.TrimSpace(loadPreferences().Model)
+}
+
+// SaveModelPreference persists the model string to preferences.yml.
+func SaveModelPreference(model string) error {
+	return savePreferences(func(p *preferences) {
+		p.Model = model
+	})
+}
+
+// ── Thinking level preference ───────────────────────────────────────────────
+
+// LoadThinkingLevelPreference reads the persisted thinking level from
+// preferences.yml. Returns "" if no preference is saved.
+func LoadThinkingLevelPreference() string {
+	return strings.TrimSpace(loadPreferences().ThinkingLevel)
+}
+
+// SaveThinkingLevelPreference persists the thinking level to preferences.yml.
+func SaveThinkingLevelPreference(level string) error {
+	return savePreferences(func(p *preferences) {
+		p.ThinkingLevel = level
+	})
+}

internal/ui/preferences_test.go

@@ -0,0 +1,180 @@
+package ui
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+)
+
+func TestSaveAndLoadThemePreference(t *testing.T) {
+	// Use a temp dir as XDG_CONFIG_HOME so we don't touch the real config.
+	tmp := t.TempDir()
+	t.Setenv("XDG_CONFIG_HOME", tmp)
+
+	// Initially no preference is saved.
+	if got := LoadThemePreference(); got != "" {
+		t.Fatalf("expected empty preference, got %q", got)
+	}
+
+	// Save a preference.
+	if err := SaveThemePreference("dracula"); err != nil {
+		t.Fatalf("SaveThemePreference: %v", err)
+	}
+
+	// Load it back.
+	if got := LoadThemePreference(); got != "dracula" {
+		t.Fatalf("expected %q, got %q", "dracula", got)
+	}
+
+	// Overwrite with a different theme.
+	if err := SaveThemePreference("nord"); err != nil {
+		t.Fatalf("SaveThemePreference: %v", err)
+	}
+	if got := LoadThemePreference(); got != "nord" {
+		t.Fatalf("expected %q, got %q", "nord", got)
+	}
+
+	// Verify the file exists and is valid YAML.
+	path := filepath.Join(tmp, "kit", "preferences.yml")
+	data, err := os.ReadFile(path)
+	if err != nil {
+		t.Fatalf("reading preferences file: %v", err)
+	}
+	if len(data) == 0 {
+		t.Fatal("preferences file is empty")
+	}
+}
+
+func TestLoadThemePreference_MissingFile(t *testing.T) {
+	tmp := t.TempDir()
+	t.Setenv("XDG_CONFIG_HOME", tmp)
+
+	// No file exists — should return empty string, not error.
+	if got := LoadThemePreference(); got != "" {
+		t.Fatalf("expected empty string for missing file, got %q", got)
+	}
+}
+
+func TestLoadThemePreference_InvalidYAML(t *testing.T) {
+	tmp := t.TempDir()
+	t.Setenv("XDG_CONFIG_HOME", tmp)
+
+	// Write invalid YAML.
+	dir := filepath.Join(tmp, "kit")
+	_ = os.MkdirAll(dir, 0o755)
+	_ = os.WriteFile(filepath.Join(dir, "preferences.yml"), []byte(":::bad yaml"), 0o644)
+
+	// Should return empty string, not panic.
+	if got := LoadThemePreference(); got != "" {
+		t.Fatalf("expected empty string for invalid YAML, got %q", got)
+	}
+}
+
+func TestSaveThemePreference_PreservesOtherFields(t *testing.T) {
+	tmp := t.TempDir()
+	t.Setenv("XDG_CONFIG_HOME", tmp)
+
+	// Pre-populate with extra content (simulating future fields).
+	dir := filepath.Join(tmp, "kit")
+	_ = os.MkdirAll(dir, 0o755)
+	_ = os.WriteFile(filepath.Join(dir, "preferences.yml"), []byte("theme: old\n"), 0o644)
+
+	// Overwrite theme.
+	if err := SaveThemePreference("catppuccin"); err != nil {
+		t.Fatalf("SaveThemePreference: %v", err)
+	}
+
+	if got := LoadThemePreference(); got != "catppuccin" {
+		t.Fatalf("expected %q, got %q", "catppuccin", got)
+	}
+}
+
+func TestSaveAndLoadModelPreference(t *testing.T) {
+	tmp := t.TempDir()
+	t.Setenv("XDG_CONFIG_HOME", tmp)
+
+	// Initially empty.
+	if got := LoadModelPreference(); got != "" {
+		t.Fatalf("expected empty, got %q", got)
+	}
+
+	// Save a model.
+	if err := SaveModelPreference("anthropic/claude-sonnet-4-5-20250929"); err != nil {
+		t.Fatalf("SaveModelPreference: %v", err)
+	}
+	if got := LoadModelPreference(); got != "anthropic/claude-sonnet-4-5-20250929" {
+		t.Fatalf("expected %q, got %q", "anthropic/claude-sonnet-4-5-20250929", got)
+	}
+
+	// Overwrite.
+	if err := SaveModelPreference("openai/gpt-4o"); err != nil {
+		t.Fatalf("SaveModelPreference: %v", err)
+	}
+	if got := LoadModelPreference(); got != "openai/gpt-4o" {
+		t.Fatalf("expected %q, got %q", "openai/gpt-4o", got)
+	}
+}
+
+func TestSaveAndLoadThinkingLevelPreference(t *testing.T) {
+	tmp := t.TempDir()
+	t.Setenv("XDG_CONFIG_HOME", tmp)
+
+	// Initially empty.
+	if got := LoadThinkingLevelPreference(); got != "" {
+		t.Fatalf("expected empty, got %q", got)
+	}
+
+	// Save a level.
+	if err := SaveThinkingLevelPreference("medium"); err != nil {
+		t.Fatalf("SaveThinkingLevelPreference: %v", err)
+	}
+	if got := LoadThinkingLevelPreference(); got != "medium" {
+		t.Fatalf("expected %q, got %q", "medium", got)
+	}
+
+	// Overwrite.
+	if err := SaveThinkingLevelPreference("high"); err != nil {
+		t.Fatalf("SaveThinkingLevelPreference: %v", err)
+	}
+	if got := LoadThinkingLevelPreference(); got != "high" {
+		t.Fatalf("expected %q, got %q", "high", got)
+	}
+}
+
+func TestPreferencesPreserveEachOther(t *testing.T) {
+	tmp := t.TempDir()
+	t.Setenv("XDG_CONFIG_HOME", tmp)
+
+	// Save all three preferences.
+	if err := SaveThemePreference("dracula"); err != nil {
+		t.Fatal(err)
+	}
+	if err := SaveModelPreference("anthropic/claude-haiku-3-5-20241022"); err != nil {
+		t.Fatal(err)
+	}
+	if err := SaveThinkingLevelPreference("high"); err != nil {
+		t.Fatal(err)
+	}
+
+	// All three should be preserved.
+	if got := LoadThemePreference(); got != "dracula" {
+		t.Fatalf("theme: expected %q, got %q", "dracula", got)
+	}
+	if got := LoadModelPreference(); got != "anthropic/claude-haiku-3-5-20241022" {
+		t.Fatalf("model: expected %q, got %q", "anthropic/claude-haiku-3-5-20241022", got)
+	}
+	if got := LoadThinkingLevelPreference(); got != "high" {
+		t.Fatalf("thinking_level: expected %q, got %q", "high", got)
+	}
+
+	// Updating one should not affect the others.
+	if err := SaveModelPreference("openai/gpt-4o"); err != nil {
+		t.Fatal(err)
+	}
+	if got := LoadThemePreference(); got != "dracula" {
+		t.Fatalf("theme after model update: expected %q, got %q", "dracula", got)
+	}
+	if got := LoadThinkingLevelPreference(); got != "high" {
+		t.Fatalf("thinking_level after model update: expected %q, got %q", "high", got)
+	}
+}

internal/ui/session_selector.go

@@ -0,0 +1,535 @@
+package ui
+
+import (
+	"fmt"
+	"regexp"
+	"strings"
+	"time"
+	"unicode/utf8"
+
+	"charm.land/bubbles/v2/key"
+	tea "charm.land/bubbletea/v2"
+	"charm.land/lipgloss/v2"
+
+	"github.com/mark3labs/kit/internal/session"
+)
+
+// SessionSelectedMsg is sent when the user selects a session from the picker.
+type SessionSelectedMsg struct {
+	Path string // absolute path to the JSONL session file
+}
+
+// SessionSelectorCancelledMsg is sent when the user cancels the picker.
+type SessionSelectorCancelledMsg struct{}
+
+// SessionDeletedMsg is sent after a session is deleted so the parent can
+// react (e.g. print a message).
+type SessionDeletedMsg struct {
+	Name string
+}
+
+// SessionScopeMode controls which sessions are shown.
+type SessionScopeMode int
+
+const (
+	SessionScopeCwd SessionScopeMode = iota // current folder only
+	SessionScopeAll                         // all sessions across projects
+)
+
+func (m SessionScopeMode) String() string {
+	if m == SessionScopeAll {
+		return "All"
+	}
+	return "Current Folder"
+}
+
+// SessionFilterMode controls filtering of the session list.
+type SessionFilterMode int
+
+const (
+	SessionFilterAll   SessionFilterMode = iota // show all sessions
+	SessionFilterNamed                          // only named sessions
+)
+
+func (m SessionFilterMode) String() string {
+	if m == SessionFilterNamed {
+		return "Named"
+	}
+	return "All"
+}
+
+// controlCharsRe matches ASCII control characters for stripping from previews.
+var controlCharsRe = regexp.MustCompile(`[\x00-\x1f\x7f]`)
+
+// SessionSelectorComponent is a full-screen Bubble Tea component that lets
+// the user browse and select from available sessions. Modeled after pi's
+// session picker: right-aligned metadata, background-highlighted selection,
+// scope/filter toggles, and inline search.
+type SessionSelectorComponent struct {
+	allSessions []session.SessionInfo
+	cwdSessions []session.SessionInfo
+	filtered    []session.SessionInfo
+
+	cursor int
+	search string
+
+	scope  SessionScopeMode
+	filter SessionFilterMode
+
+	// currentPath is the active session file path for marking it in the list.
+	currentPath string
+
+	width  int
+	height int
+	active bool
+
+	// confirmDelete is non-negative when a delete confirmation is pending.
+	confirmDelete int
+}
+
+// NewSessionSelector creates a session selector. It loads sessions for the
+// current working directory and all sessions across projects. If cwd is
+// empty, only "All" scope is available.
+func NewSessionSelector(cwd string, width, height int) *SessionSelectorComponent {
+	ss := &SessionSelectorComponent{
+		width:         width,
+		height:        height,
+		active:        true,
+		confirmDelete: -1,
+	}
+
+	// Load sessions (errors are swallowed — empty list is fine).
+	if cwd != "" {
+		ss.cwdSessions, _ = session.ListSessions(cwd)
+		ss.scope = SessionScopeCwd
+	}
+	ss.allSessions, _ = session.ListAllSessions()
+
+	if cwd == "" || len(ss.cwdSessions) == 0 {
+		ss.scope = SessionScopeAll
+	}
+
+	ss.rebuildFiltered()
+	return ss
+}
+
+// SetCurrentPath sets the currently active session path so the picker can
+// highlight it in the list.
+func (ss *SessionSelectorComponent) SetCurrentPath(path string) {
+	ss.currentPath = path
+}
+
+// Init implements tea.Model.
+func (ss *SessionSelectorComponent) Init() tea.Cmd {
+	return nil
+}
+
+// Update implements tea.Model.
+func (ss *SessionSelectorComponent) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
+	switch msg := msg.(type) {
+	case tea.WindowSizeMsg:
+		ss.width = msg.Width
+		ss.height = msg.Height
+		return ss, nil
+
+	case tea.KeyPressMsg:
+		// Delete confirmation mode.
+		if ss.confirmDelete >= 0 {
+			switch msg.String() {
+			case "y", "Y":
+				idx := ss.confirmDelete
+				ss.confirmDelete = -1
+				if idx < len(ss.filtered) {
+					info := ss.filtered[idx]
+					if err := session.DeleteSession(info.Path); err == nil {
+						name := sessionDisplayName(info)
+						ss.removeSession(info.Path)
+						ss.rebuildFiltered()
+						return ss, func() tea.Msg {
+							return SessionDeletedMsg{Name: name}
+						}
+					}
+				}
+				return ss, nil
+			default:
+				ss.confirmDelete = -1
+				return ss, nil
+			}
+		}
+
+		switch {
+		case key.Matches(msg, key.NewBinding(key.WithKeys("up", "k"))):
+			if ss.cursor > 0 {
+				ss.cursor--
+			}
+
+		case key.Matches(msg, key.NewBinding(key.WithKeys("down", "j"))):
+			if ss.cursor < len(ss.filtered)-1 {
+				ss.cursor++
+			}
+
+		case key.Matches(msg, key.NewBinding(key.WithKeys("pgup"))):
+			ss.cursor -= ss.visibleHeight()
+			if ss.cursor < 0 {
+				ss.cursor = 0
+			}
+
+		case key.Matches(msg, key.NewBinding(key.WithKeys("pgdown"))):
+			ss.cursor += ss.visibleHeight()
+			if ss.cursor >= len(ss.filtered) {
+				ss.cursor = len(ss.filtered) - 1
+			}
+			if ss.cursor < 0 {
+				ss.cursor = 0
+			}
+
+		case key.Matches(msg, key.NewBinding(key.WithKeys("home"))):
+			ss.cursor = 0
+
+		case key.Matches(msg, key.NewBinding(key.WithKeys("end"))):
+			ss.cursor = max(len(ss.filtered)-1, 0)
+
+		case key.Matches(msg, key.NewBinding(key.WithKeys("enter"))):
+			if ss.cursor < len(ss.filtered) {
+				info := ss.filtered[ss.cursor]
+				ss.active = false
+				return ss, func() tea.Msg {
+					return SessionSelectedMsg{Path: info.Path}
+				}
+			}
+
+		case key.Matches(msg, key.NewBinding(key.WithKeys("esc"))):
+			if ss.search != "" {
+				ss.search = ""
+				ss.rebuildFiltered()
+			} else {
+				ss.active = false
+				return ss, func() tea.Msg {
+					return SessionSelectorCancelledMsg{}
+				}
+			}
+
+		case key.Matches(msg, key.NewBinding(key.WithKeys("tab"))):
+			if ss.scope == SessionScopeCwd {
+				ss.scope = SessionScopeAll
+			} else {
+				ss.scope = SessionScopeCwd
+			}
+			ss.rebuildFiltered()
+
+		case key.Matches(msg, key.NewBinding(key.WithKeys("ctrl+n"))):
+			if ss.filter == SessionFilterAll {
+				ss.filter = SessionFilterNamed
+			} else {
+				ss.filter = SessionFilterAll
+			}
+			ss.rebuildFiltered()
+
+		case key.Matches(msg, key.NewBinding(key.WithKeys("d"))):
+			if ss.cursor < len(ss.filtered) {
+				ss.confirmDelete = ss.cursor
+			}
+			return ss, nil
+
+		default:
+			if msg.Text != "" && len(msg.Text) == 1 {
+				ch := msg.Text[0]
+				if ch >= 32 && ch < 127 {
+					ss.search += string(ch)
+					ss.rebuildFiltered()
+				}
+			}
+			if key.Matches(msg, key.NewBinding(key.WithKeys("backspace"))) && len(ss.search) > 0 {
+				ss.search = ss.search[:len(ss.search)-1]
+				ss.rebuildFiltered()
+			}
+		}
+	}
+	return ss, nil
+}
+
+// View implements tea.Model.
+func (ss *SessionSelectorComponent) View() tea.View {
+	theme := GetTheme()
+	w := ss.width
+	var b strings.Builder
+
+	// ── Header: title + scope badges ─────────────────────────────
+	titleStyle := lipgloss.NewStyle().Bold(true).Foreground(theme.Accent).PaddingLeft(1)
+	b.WriteString(titleStyle.Render(fmt.Sprintf("Resume Session (%s)", ss.scope)))
+	b.WriteString("\n")
+
+	// ── Help / keybindings ───────────────────────────────────────
+	helpStyle := lipgloss.NewStyle().Foreground(theme.Muted).PaddingLeft(1)
+	if w >= 75 {
+		b.WriteString(helpStyle.Render("tab: scope  N: named  D: delete  R: rename  type to search  esc: cancel"))
+	} else if w >= 50 {
+		b.WriteString(helpStyle.Render("tab scope  N named  D del  type to search  esc"))
+	} else {
+		b.WriteString(helpStyle.Render("tab N D esc"))
+	}
+	b.WriteString("\n")
+
+	// ── Search (only shown when active) ──────────────────────────
+	if ss.search != "" {
+		searchStyle := lipgloss.NewStyle().Foreground(theme.Info).PaddingLeft(1)
+		b.WriteString(searchStyle.Render(fmt.Sprintf("> %s", ss.search)))
+		b.WriteString("\n")
+	}
+
+	b.WriteString("\n")
+
+	// ── Delete confirmation ──────────────────────────────────────
+	if ss.confirmDelete >= 0 && ss.confirmDelete < len(ss.filtered) {
+		warnStyle := lipgloss.NewStyle().Foreground(theme.Error).Bold(true).PaddingLeft(1)
+		name := sessionDisplayName(ss.filtered[ss.confirmDelete])
+		b.WriteString(warnStyle.Render(fmt.Sprintf("Delete %q? (y/N)", truncateRunes(name, 40))))
+		b.WriteString("\n")
+	}
+
+	// ── Session list ─────────────────────────────────────────────
+	if len(ss.filtered) == 0 {
+		emptyStyle := lipgloss.NewStyle().Foreground(theme.Muted).PaddingLeft(2)
+		if ss.search != "" {
+			b.WriteString(emptyStyle.Render(fmt.Sprintf("No sessions matching %q", ss.search)))
+		} else if ss.filter == SessionFilterNamed {
+			b.WriteString(emptyStyle.Render("No named sessions. Press N to show all."))
+		} else if ss.scope == SessionScopeCwd {
+			b.WriteString(emptyStyle.Render("No sessions in current folder. Press tab to view all."))
+		} else {
+			b.WriteString(emptyStyle.Render("No sessions found"))
+		}
+		b.WriteString("\n")
+	} else {
+		visH := ss.visibleHeight()
+
+		// Center the cursor in the visible window.
+		startIdx := max(0, min(ss.cursor-visH/2, len(ss.filtered)-visH))
+		endIdx := min(startIdx+visH, len(ss.filtered))
+
+		for i := startIdx; i < endIdx; i++ {
+			info := ss.filtered[i]
+			isCursor := i == ss.cursor
+			isCurrent := info.Path == ss.currentPath
+			isDeleting := i == ss.confirmDelete
+			line := ss.renderEntry(info, isCursor, isCurrent, isDeleting, w)
+			b.WriteString(line)
+			b.WriteString("\n")
+		}
+
+		// Scroll position indicator.
+		if len(ss.filtered) > visH {
+			posStyle := lipgloss.NewStyle().Foreground(theme.Muted).PaddingLeft(2)
+			b.WriteString(posStyle.Render(fmt.Sprintf("(%d/%d)", ss.cursor+1, len(ss.filtered))))
+			b.WriteString("\n")
+		}
+	}
+
+	return tea.NewView(b.String())
+}
+
+// IsActive returns whether the selector is still accepting input.
+func (ss *SessionSelectorComponent) IsActive() bool {
+	return ss.active
+}
+
+// --- Internal helpers ---
+
+func (ss *SessionSelectorComponent) visibleHeight() int {
+	// Reserve: title(1) + help(1) + blank(1) + scroll indicator(1) = 4.
+	// Optional: search(1), delete confirm(1).
+	chrome := 4
+	if ss.search != "" {
+		chrome++
+	}
+	if ss.confirmDelete >= 0 {
+		chrome++
+	}
+	return max(ss.height-chrome, 3)
+}
+
+func (ss *SessionSelectorComponent) rebuildFiltered() {
+	var source []session.SessionInfo
+	if ss.scope == SessionScopeCwd {
+		source = ss.cwdSessions
+	} else {
+		source = ss.allSessions
+	}
+
+	if ss.filter == SessionFilterNamed {
+		var named []session.SessionInfo
+		for _, s := range source {
+			if s.Name != "" {
+				named = append(named, s)
+			}
+		}
+		source = named
+	}
+
+	if ss.search != "" {
+		query := strings.ToLower(ss.search)
+		var matches []session.SessionInfo
+		for _, s := range source {
+			haystack := strings.ToLower(s.Name + " " + s.FirstMessage + " " + s.Cwd)
+			if strings.Contains(haystack, query) {
+				matches = append(matches, s)
+			}
+		}
+		ss.filtered = matches
+	} else {
+		ss.filtered = source
+	}
+
+	if ss.cursor >= len(ss.filtered) {
+		ss.cursor = max(len(ss.filtered)-1, 0)
+	}
+}
+
+func (ss *SessionSelectorComponent) removeSession(path string) {
+	ss.cwdSessions = removeByPath(ss.cwdSessions, path)
+	ss.allSessions = removeByPath(ss.allSessions, path)
+}
+
+func removeByPath(sessions []session.SessionInfo, path string) []session.SessionInfo {
+	result := make([]session.SessionInfo, 0, len(sessions))
+	for _, s := range sessions {
+		if s.Path != path {
+			result = append(result, s)
+		}
+	}
+	return result
+}
+
+// renderEntry renders a single session line with right-aligned metadata.
+// Layout: [cursor 2] [message ...variable...] [padding] [count age] [cwd?]
+func (ss *SessionSelectorComponent) renderEntry(info session.SessionInfo, isCursor, isCurrent, isDeleting bool, width int) string {
+	theme := GetTheme()
+
+	// ── Cursor indicator (2 chars) ───────────────────────────────
+	cursorStr := "  "
+	if isCursor {
+		cursorStr = lipgloss.NewStyle().Foreground(theme.Accent).Render("› ")
+	}
+	const cursorW = 2
+
+	// ── Right part: message count + relative time (+ optional cwd) ──
+	age := relativeTime(info.Modified)
+	msgCount := fmt.Sprintf("%d", info.MessageCount)
+	rightPart := msgCount + " " + age
+	if ss.scope == SessionScopeAll && info.Cwd != "" {
+		shortCwd := shortenPath(info.Cwd)
+		if len(shortCwd) > 25 {
+			shortCwd = "..." + shortCwd[len(shortCwd)-22:]
+		}
+		rightPart = shortCwd + " " + rightPart
+	}
+	rightW := utf8.RuneCountInString(rightPart)
+
+	// ── Message text ─────────────────────────────────────────────
+	displayText := sessionDisplayName(info)
+	// Strip control characters and collapse whitespace.
+	displayText = controlCharsRe.ReplaceAllString(displayText, " ")
+	displayText = strings.Join(strings.Fields(displayText), " ")
+
+	availableForMsg := max(width-cursorW-rightW-2, 10) // 2 for min spacing
+	displayText = truncateRunes(displayText, availableForMsg)
+	msgW := utf8.RuneCountInString(displayText)
+
+	// ── Style the message ────────────────────────────────────────
+	msgStyle := lipgloss.NewStyle()
+	switch {
+	case isDeleting:
+		msgStyle = msgStyle.Foreground(theme.Error)
+	case isCurrent:
+		msgStyle = msgStyle.Foreground(theme.Accent)
+	case info.Name != "":
+		msgStyle = msgStyle.Foreground(theme.Warning)
+	default:
+		msgStyle = msgStyle.Foreground(theme.Text)
+	}
+	if isCursor {
+		msgStyle = msgStyle.Bold(true)
+	}
+
+	styledMsg := msgStyle.Render(displayText)
+
+	// ── Style the right part ─────────────────────────────────────
+	rightColor := theme.Muted
+	if isDeleting {
+		rightColor = theme.Error
+	}
+	styledRight := lipgloss.NewStyle().Foreground(rightColor).Render(rightPart)
+
+	// ── Assemble with spacing ────────────────────────────────────
+	spacing := max(width-cursorW-msgW-rightW, 1)
+
+	line := cursorStr + styledMsg + strings.Repeat(" ", spacing) + styledRight
+
+	// ── Background highlight for selected row ────────────────────
+	if isCursor {
+		// Use a subtle background highlight. We apply it by wrapping the
+		// full line in a style with a background color.
+		bgStyle := lipgloss.NewStyle().
+			Background(theme.Highlight).
+			Width(width)
+		line = bgStyle.Render(line)
+	}
+
+	return line
+}
+
+// --- Package helpers ---
+
+// sessionDisplayName returns the best display string for a session:
+// the name if set, the first message, or a fallback.
+func sessionDisplayName(info session.SessionInfo) string {
+	if info.Name != "" {
+		return info.Name
+	}
+	if info.FirstMessage != "" {
+		return info.FirstMessage
+	}
+	return "(empty session)"
+}
+
+// truncateRunes truncates a string to at most maxRunes runes, appending "..."
+// if truncated.
+func truncateRunes(s string, maxRunes int) string {
+	if maxRunes <= 0 {
+		return ""
+	}
+	runes := []rune(s)
+	if len(runes) <= maxRunes {
+		return s
+	}
+	if maxRunes <= 3 {
+		return string(runes[:maxRunes])
+	}
+	return string(runes[:maxRunes-1]) + "…"
+}
+
+// shortenPath replaces the user's home directory prefix with ~.
+func shortenPath(path string) string {
+	return tildeHome(path)
+}
+
+// relativeTime formats a time as a short relative string like "5m", "2h", "3d".
+func relativeTime(t time.Time) string {
+	d := time.Since(t)
+	switch {
+	case d < time.Minute:
+		return "now"
+	case d < time.Hour:
+		return fmt.Sprintf("%dm", int(d.Minutes()))
+	case d < 24*time.Hour:
+		return fmt.Sprintf("%dh", int(d.Hours()))
+	case d < 7*24*time.Hour:
+		return fmt.Sprintf("%dd", int(d.Hours()/24))
+	case d < 30*24*time.Hour:
+		return fmt.Sprintf("%dw", int(d.Hours()/(24*7)))
+	case d < 365*24*time.Hour:
+		return fmt.Sprintf("%dmo", int(d.Hours()/(24*30)))
+	default:
+		return fmt.Sprintf("%dy", int(d.Hours()/(24*365)))
+	}
+}

internal/ui/stream.go

@@ -484,6 +484,7 @@ func (s *StreamComponent) renderReasoningBlock(reasoning string) string {
 
 	contentStyle := lipgloss.NewStyle().
 		Foreground(theme.Muted).
+		Background(theme.MutedBorder).
 		Italic(true)
 
 	var parts []string
@@ -495,6 +496,7 @@ func (s *StreamComponent) renderReasoningBlock(reasoning string) string {
 		hidden := len(lines) - maxCollapsedLines
 		hintStyle := lipgloss.NewStyle().
 			Foreground(theme.VeryMuted).
+			Background(theme.MutedBorder).
 			Italic(true)
 		parts = append(parts, hintStyle.Render(fmt.Sprintf("... (%d lines hidden)", hidden)))
 		lines = lines[len(lines)-maxCollapsedLines:]
@@ -517,8 +519,8 @@ func (s *StreamComponent) renderReasoningBlock(reasoning string) string {
 		} else {
 			durationStr = fmt.Sprintf("%.1fs", duration.Seconds())
 		}
-		footer := lipgloss.NewStyle().Foreground(theme.VeryMuted).Render("Thought for ") +
-			lipgloss.NewStyle().Foreground(theme.Info).Render(durationStr)
+		footer := lipgloss.NewStyle().Foreground(theme.VeryMuted).Background(theme.MutedBorder).Render("Thought for ") +
+			lipgloss.NewStyle().Foreground(theme.Info).Background(theme.MutedBorder).Render(durationStr)
 		parts = append(parts, footer)
 	}
 

internal/ui/themes.go

@@ -439,7 +439,23 @@ func LoadThemeByName(name string) (Theme, error) {
 }
 
 // ApplyTheme loads a theme by name and sets it as the active global theme.
+// The selection is persisted to ~/.config/kit/preferences.yml so it survives
+// across sessions. Persistence errors are silently ignored — the theme is
+// still applied in-memory even if the write fails.
 func ApplyTheme(name string) error {
+	t, err := LoadThemeByName(name)
+	if err != nil {
+		return err
+	}
+	SetTheme(t)
+	_ = SaveThemePreference(name)
+	return nil
+}
+
+// ApplyThemeWithoutSave loads a theme by name and sets it as the active global
+// theme without persisting the choice. Used at startup to restore a previously
+// saved preference without redundantly re-writing it.
+func ApplyThemeWithoutSave(name string) error {
 	t, err := LoadThemeByName(name)
 	if err != nil {
 		return err

internal/ui/tool_renderers.go

@@ -14,6 +14,7 @@ import (
 	"github.com/alecthomas/chroma/v2/lexers"
 	"github.com/alecthomas/chroma/v2/styles"
 	udiff "github.com/aymanbagabas/go-udiff"
+	xansi "github.com/charmbracelet/x/ansi"
 )
 
 // Maximum visible lines per tool type before truncation.
@@ -228,7 +229,7 @@ func renderDiffBlock(before, after string, startLine int, width int) string {
 	gutterMissing := lipgloss.NewStyle().Background(theme.DiffMissingBg)
 
 	contentInsert := lipgloss.NewStyle().Background(theme.DiffInsertBg)
-	contentDelete := lipgloss.NewStyle().Background(theme.DiffDeleteBg).Strikethrough(true)
+	contentDelete := lipgloss.NewStyle().Background(theme.DiffDeleteBg)
 	contentEqual := lipgloss.NewStyle().Foreground(theme.Muted).Background(theme.DiffEqualBg)
 	contentMissing := lipgloss.NewStyle().Background(theme.DiffMissingBg)
 
@@ -322,6 +323,8 @@ func renderLsBody(toolResult string, width int) string {
 
 	var result []string
 	for _, line := range lines {
+		// Truncate before styling to prevent wrapping.
+		line = truncateLine(line, codeWidth-1) // account for PaddingLeft(1)
 		styled := codeStyle.Width(codeWidth).Render(line)
 		result = append(result, indent+styled)
 	}
@@ -431,7 +434,8 @@ func renderCodeBlock(content, fileName string, width int) string {
 		// If this line has no line number, it's a metadata/footer line (e.g. truncation notice).
 		if p.lineNum == "" {
 			// Render footer lines with code background but no gutter
-			footer := codeStyle.Width(codeWidth).Render(p.code)
+			truncatedFooter := truncateLine(p.code, codeWidth-1) // account for PaddingLeft(1)
+			footer := codeStyle.Width(codeWidth).Render(truncatedFooter)
 			emptyGutter := gutterStyle.Width(gutterWidth).Render("")
 			result = append(result, codeIndent+lipgloss.JoinHorizontal(lipgloss.Top, emptyGutter, footer))
 			continue
@@ -445,6 +449,9 @@ func renderCodeBlock(content, fileName string, width int) string {
 		} else {
 			codePart = p.code
 		}
+		// Truncate the (possibly ANSI-highlighted) line to fit within
+		// the code column, preventing lipgloss from wrapping it.
+		codePart = truncateLine(codePart, codeWidth-1) // account for PaddingLeft(1)
 		styledCode := codeStyle.Width(codeWidth).Render(codePart)
 
 		result = append(result, codeIndent+lipgloss.JoinHorizontal(lipgloss.Top, gutter, styledCode))
@@ -528,6 +535,9 @@ func renderWriteBlock(content, fileName string, width int) string {
 		} else {
 			codePart = line
 		}
+		// Truncate the (possibly ANSI-highlighted) line to fit within
+		// the code column, preventing lipgloss from wrapping it.
+		codePart = truncateLine(codePart, codeWidth-1) // account for PaddingLeft(1)
 		styledCode := writeStyle.Width(codeWidth).Render(codePart)
 
 		result = append(result, codeIndent+lipgloss.JoinHorizontal(lipgloss.Top, gutter, styledCode))
@@ -578,9 +588,16 @@ func renderBashBody(toolResult string, width int) string {
 	}
 
 	const lineIndent = "  "
+	// Truncate individual lines to the available width so they never wrap.
+	// This mirrors Crush's approach: truncate, don't wrap.
+	lineWidth := max(width-len(lineIndent), 20)
+	// Account for PaddingLeft(1) on the output/stderr styles
+	maxLineChars := lineWidth - 1
+
 	var rendered []string
 	inStderr := false
 	for _, line := range lines {
+		line = truncateLine(line, maxLineChars)
 		// Detect the STDERR: label that Kit's bash tool emits
 		if strings.TrimSpace(line) == "STDERR:" {
 			inStderr = true
@@ -682,23 +699,28 @@ func syntaxHighlight(source, fileName string) string {
 // Helpers
 // ---------------------------------------------------------------------------
 
-// padRight pads s with spaces to exactly width characters.
+// padRight pads s with spaces to exactly width visual characters.
+// This is ANSI-aware: it measures the visual width of s (ignoring escape
+// codes and accounting for wide characters) before padding or truncating.
 func padRight(s string, width int) string {
-	if len(s) >= width {
-		return s[:width]
+	w := xansi.StringWidth(s)
+	if w >= width {
+		return xansi.Truncate(s, width, "")
 	}
-	return s + strings.Repeat(" ", width-len(s))
+	return s + strings.Repeat(" ", width-w)
 }
 
-// truncateLine truncates a line to maxWidth, adding "…" if truncated.
+// truncateLine truncates a line to maxWidth visual characters, adding "…"
+// if truncated. This is ANSI-aware: escape codes are preserved and wide
+// characters are measured correctly.
 func truncateLine(s string, maxWidth int) string {
-	if len(s) <= maxWidth {
+	if xansi.StringWidth(s) <= maxWidth {
 		return s
 	}
 	if maxWidth < 2 {
-		return s[:maxWidth]
+		return xansi.Truncate(s, maxWidth, "")
 	}
-	return s[:maxWidth-1] + "…"
+	return xansi.Truncate(s, maxWidth, "…")
 }
 
 // ---------------------------------------------------------------------------
@@ -858,12 +880,10 @@ func renderBashCompact(toolResult string, width int) string {
 		display = display[:maxLines]
 	}
 
-	// Truncate each line to available width
+	// Truncate each line to available width (ANSI-aware)
 	lineMax := max(width-4, 20)
 	for i, line := range display {
-		if len(line) > lineMax {
-			display[i] = line[:lineMax-3] + "..."
-		}
+		display[i] = truncateLine(line, lineMax)
 	}
 
 	summary := strings.Join(display, "\n")
@@ -940,10 +960,8 @@ func extractSubagentPreview(content string, maxLines, maxWidth int) string {
 			continue
 		}
 
-		// Truncate long lines
-		if len(trimmed) > maxWidth {
-			trimmed = trimmed[:maxWidth-3] + "..."
-		}
+		// Truncate long lines (ANSI-aware)
+		trimmed = truncateLine(trimmed, maxWidth)
 		preview = append(preview, trimmed)
 
 		if len(preview) >= maxLines {

internal/ui/tree_selector.go

@@ -430,6 +430,8 @@ func (ts *TreeSelectorComponent) renderNode(node FlatNode, isCursor, isLeaf bool
 		}
 	case *session.BranchSummaryEntry:
 		style = lipgloss.NewStyle().Foreground(theme.Warning).Italic(true)
+	case *session.CompactionEntry:
+		style = lipgloss.NewStyle().Foreground(theme.Info).Italic(true)
 	default:
 		style = lipgloss.NewStyle().Foreground(theme.Muted)
 	}
@@ -483,6 +485,13 @@ func (ts *TreeSelectorComponent) entryDisplayText(entry any) string {
 		}
 		return fmt.Sprintf("branch summary: %s", summary)
 
+	case *session.CompactionEntry:
+		summary := e.Summary
+		if len(summary) > 60 {
+			summary = summary[:60] + "..."
+		}
+		return fmt.Sprintf("compaction: %s", summary)
+
 	case *session.LabelEntry:
 		return fmt.Sprintf("label: %s", e.Label)
 

pkg/kit/compaction.go

@@ -4,6 +4,8 @@ import (
 	"context"
 	"fmt"
 
+	"charm.land/fantasy"
+
 	"github.com/mark3labs/kit/internal/compaction"
 )
 
@@ -83,8 +85,10 @@ func (m *Kit) GetContextStats() ContextStats {
 // customInstructions is optional text appended to the summary prompt (e.g.
 // "Focus on the API design decisions"). Pass "" for the default prompt.
 //
-// After compaction, the tree session is cleared and replaced with the
-// compacted messages (summary + preserved recent messages).
+// Compaction is non-destructive: a CompactionEntry is appended to the session
+// tree recording the summary and the first kept entry ID. Old messages remain
+// on disk but are skipped when building the LLM context — the summary is
+// injected in their place.
 func (m *Kit) Compact(ctx context.Context, opts *CompactionOptions, customInstructions string) (*CompactionResult, error) {
 	return m.compactInternal(ctx, opts, customInstructions, false)
 }
@@ -112,7 +116,7 @@ func (m *Kit) compactInternal(ctx context.Context, opts *CompactionOptions, cust
 		return nil, fmt.Errorf("cannot compact: need at least 2 messages")
 	}
 
-	// Run before-compact hook — extensions can cancel compaction.
+	// Run before-compact hook — extensions can cancel or provide a custom summary.
 	if m.beforeCompact.hasHooks() {
 		stats := m.GetContextStats()
 		if hookResult := m.beforeCompact.run(BeforeCompactHook{
@@ -121,17 +125,32 @@ func (m *Kit) compactInternal(ctx context.Context, opts *CompactionOptions, cust
 			UsagePercent:    stats.UsagePercent,
 			MessageCount:    stats.MessageCount,
 			IsAutomatic:     isAutomatic,
-		}); hookResult != nil && hookResult.Cancel {
-			reason := hookResult.Reason
-			if reason == "" {
-				reason = "compaction cancelled by extension"
+		}); hookResult != nil {
+			if hookResult.Cancel {
+				reason := hookResult.Reason
+				if reason == "" {
+					reason = "compaction cancelled by extension"
+				}
+				return nil, fmt.Errorf("%s", reason)
 			}
-			return nil, fmt.Errorf("%s", reason)
+			// Extension provided a custom summary — use it directly.
+			if hookResult.Summary != "" {
+				return m.applyCustomCompaction(hookResult.Summary, messages, opts)
+			}
+		}
+	}
+
+	// Carry forward file tracking from previous compaction.
+	var prev *compaction.PreviousCompaction
+	if lastCompaction := m.treeSession.GetLastCompaction(); lastCompaction != nil {
+		prev = &compaction.PreviousCompaction{
+			ReadFiles:     lastCompaction.ReadFiles,
+			ModifiedFiles: lastCompaction.ModifiedFiles,
 		}
 	}
 
 	model := m.agent.GetModel()
-	result, newMessages, err := compaction.Compact(ctx, model, messages, *opts, customInstructions)
+	result, _, err := compaction.Compact(ctx, model, messages, *opts, customInstructions, prev)
 	if err != nil {
 		return nil, err
 	}
@@ -139,11 +158,82 @@ func (m *Kit) compactInternal(ctx context.Context, opts *CompactionOptions, cust
 		return nil, nil
 	}
 
-	// Replace the session contents with the compacted messages.
-	// Reset the tree leaf and re-add the compacted messages.
-	m.treeSession.ResetLeaf()
-	if err := m.treeSession.AddFantasyMessages(newMessages); err != nil {
-		return nil, fmt.Errorf("failed to persist compacted messages: %w", err)
+	// Non-destructive: append a CompactionEntry to the session tree instead
+	// of clearing and rewriting messages.
+	entryIDs := m.treeSession.GetContextEntryIDs()
+	firstKeptEntryID := ""
+	if result.CutPoint >= 0 && result.CutPoint < len(entryIDs) {
+		firstKeptEntryID = entryIDs[result.CutPoint]
+	}
+
+	if _, err := m.treeSession.AppendCompaction(
+		result.Summary,
+		firstKeptEntryID,
+		result.OriginalTokens,
+		result.CompactedTokens,
+		result.MessagesRemoved,
+		result.ReadFiles,
+		result.ModifiedFiles,
+	); err != nil {
+		return nil, fmt.Errorf("failed to persist compaction entry: %w", err)
+	}
+
+	m.events.emit(CompactionEvent{
+		Summary:         result.Summary,
+		OriginalTokens:  result.OriginalTokens,
+		CompactedTokens: result.CompactedTokens,
+		MessagesRemoved: result.MessagesRemoved,
+		ReadFiles:       result.ReadFiles,
+		ModifiedFiles:   result.ModifiedFiles,
+	})
+
+	return result, nil
+}
+
+// applyCustomCompaction handles compaction when an extension provides a
+// custom summary. It still determines the cut point and persists a
+// CompactionEntry.
+func (m *Kit) applyCustomCompaction(summary string, messages []fantasy.Message, opts *CompactionOptions) (*CompactionResult, error) {
+	originalTokens := compaction.EstimateMessageTokens(messages)
+
+	cutPoint := compaction.FindCutPoint(messages, opts.KeepRecentTokens)
+	if cutPoint == 0 {
+		cutPoint = len(messages) - 1
+		if cutPoint < 1 {
+			return nil, nil
+		}
+	}
+
+	entryIDs := m.treeSession.GetContextEntryIDs()
+	firstKeptEntryID := ""
+	if cutPoint >= 0 && cutPoint < len(entryIDs) {
+		firstKeptEntryID = entryIDs[cutPoint]
+	}
+
+	// Estimate new token count.
+	summaryTokens := compaction.EstimateMessageTokens([]fantasy.Message{{
+		Role:    "system",
+		Content: []fantasy.MessagePart{fantasy.TextPart{Text: summary}},
+	}})
+	recentTokens := compaction.EstimateMessageTokens(messages[cutPoint:])
+	compactedTokens := summaryTokens + recentTokens
+
+	if _, err := m.treeSession.AppendCompaction(
+		summary,
+		firstKeptEntryID,
+		originalTokens,
+		compactedTokens,
+		cutPoint,
+		nil, nil, // no file tracking for custom summaries
+	); err != nil {
+		return nil, fmt.Errorf("failed to persist compaction entry: %w", err)
+	}
+
+	result := &CompactionResult{
+		Summary:         summary,
+		OriginalTokens:  originalTokens,
+		CompactedTokens: compactedTokens,
+		MessagesRemoved: cutPoint,
 	}
 
 	m.events.emit(CompactionEvent{

pkg/kit/events.go

@@ -39,6 +39,8 @@ const (
 	EventCompaction EventType = "compaction"
 	// EventReasoningDelta fires for each streaming reasoning/thinking chunk.
 	EventReasoningDelta EventType = "reasoning_delta"
+	// EventToolOutput fires when a tool produces streaming output chunks.
+	EventToolOutput EventType = "tool_output"
 )
 
 // ---------------------------------------------------------------------------
@@ -143,6 +145,17 @@ type ReasoningDeltaEvent struct {
 // EventType implements Event.
 func (e ReasoningDeltaEvent) EventType() EventType { return EventReasoningDelta }
 
+// ToolOutputEvent fires when a tool produces streaming output chunks (e.g., bash output).
+type ToolOutputEvent struct {
+	ToolCallID string
+	ToolName   string
+	Chunk      string
+	IsStderr   bool
+}
+
+// EventType implements Event.
+func (e ToolOutputEvent) EventType() EventType { return EventToolOutput }
+
 // MessageEndEvent fires when the assistant message is complete.
 type MessageEndEvent struct {
 	Content string
@@ -242,6 +255,8 @@ type CompactionEvent struct {
 	OriginalTokens  int
 	CompactedTokens int
 	MessagesRemoved int
+	ReadFiles       []string
+	ModifiedFiles   []string
 }
 
 // EventType implements Event.
@@ -320,6 +335,16 @@ func (m *Kit) OnToolResult(handler func(ToolResultEvent)) func() {
 	})
 }
 
+// OnToolOutput registers a handler that fires only for ToolOutputEvent
+// (streaming tool output chunks, e.g., from bash). Returns an unsubscribe function.
+func (m *Kit) OnToolOutput(handler func(ToolOutputEvent)) func() {
+	return m.Subscribe(func(e Event) {
+		if to, ok := e.(ToolOutputEvent); ok {
+			handler(to)
+		}
+	})
+}
+
 // OnStreaming registers a handler that fires only for MessageUpdateEvent
 // (streaming text chunks). Returns an unsubscribe function.
 func (m *Kit) OnStreaming(handler func(MessageUpdateEvent)) func() {
@@ -359,3 +384,78 @@ func (m *Kit) OnTurnEnd(handler func(TurnEndEvent)) func() {
 		}
 	})
 }
+
+// ---------------------------------------------------------------------------
+// Subagent event subscriptions
+// ---------------------------------------------------------------------------
+
+// subagentListenerSet holds per-tool-call listeners for subagent events.
+type subagentListenerSet struct {
+	mu        sync.RWMutex
+	listeners map[int]EventListener
+	nextID    int
+}
+
+func newSubagentListenerSet() *subagentListenerSet {
+	return &subagentListenerSet{listeners: make(map[int]EventListener)}
+}
+
+func (s *subagentListenerSet) add(listener EventListener) func() {
+	s.mu.Lock()
+	id := s.nextID
+	s.nextID++
+	s.listeners[id] = listener
+	s.mu.Unlock()
+	return func() {
+		s.mu.Lock()
+		delete(s.listeners, id)
+		s.mu.Unlock()
+	}
+}
+
+func (s *subagentListenerSet) emit(event Event) {
+	s.mu.RLock()
+	snapshot := make([]EventListener, 0, len(s.listeners))
+	for _, l := range s.listeners {
+		snapshot = append(snapshot, l)
+	}
+	s.mu.RUnlock()
+	for _, l := range snapshot {
+		l(event)
+	}
+}
+
+// SubscribeSubagent registers a listener for real-time events from a subagent
+// identified by its tool call ID. Returns an unsubscribe function.
+//
+// The listener receives the same event types as Subscribe() (ToolCallEvent,
+// MessageUpdateEvent, etc.) but scoped to the child agent's activity. If the
+// tool call ID doesn't correspond to an active or future spawn_subagent call,
+// the listener simply never fires.
+//
+// Typical usage — register inside an OnToolCall handler:
+//
+//	kit.OnToolCall(func(e kit.ToolCallEvent) {
+//	    if e.ToolName == "spawn_subagent" {
+//	        kit.SubscribeSubagent(e.ToolCallID, func(child kit.Event) {
+//	            // real-time subagent events
+//	        })
+//	    }
+//	})
+func (m *Kit) SubscribeSubagent(toolCallID string, listener EventListener) func() {
+	actual, _ := m.subagentListeners.LoadOrStore(toolCallID, newSubagentListenerSet())
+	return actual.(*subagentListenerSet).add(listener)
+}
+
+// getSubagentListenerSet returns the listener set for a tool call, or nil.
+func (m *Kit) getSubagentListenerSet(toolCallID string) *subagentListenerSet {
+	if v, ok := m.subagentListeners.Load(toolCallID); ok {
+		return v.(*subagentListenerSet)
+	}
+	return nil
+}
+
+// cleanupSubagentListeners removes the listener set for a completed tool call.
+func (m *Kit) cleanupSubagentListeners(toolCallID string) {
+	m.subagentListeners.Delete(toolCallID)
+}

pkg/kit/extensions_bridge.go

@@ -86,6 +86,20 @@ func (m *Kit) bridgeExtensions(runner *extensions.Runner) {
 		})
 	}
 
+	// Tool output streaming events (observation only).
+	if runner.HasHandlers(extensions.ToolOutput) {
+		m.Subscribe(func(e Event) {
+			if ev, ok := e.(ToolOutputEvent); ok {
+				_, _ = runner.Emit(extensions.ToolOutputEvent{
+					ToolCallID: ev.ToolCallID,
+					ToolName:   ev.ToolName,
+					Chunk:      ev.Chunk,
+					IsStderr:   ev.IsStderr,
+				})
+			}
+		})
+	}
+
 	if runner.HasHandlers(extensions.AgentEnd) {
 		m.Subscribe(func(e Event) {
 			if ev, ok := e.(TurnEndEvent); ok {
@@ -173,10 +187,17 @@ func (m *Kit) bridgeExtensions(runner *extensions.Runner) {
 				MessageCount:    h.MessageCount,
 				IsAutomatic:     h.IsAutomatic,
 			})
-			if r, ok := result.(extensions.BeforeCompactResult); ok && r.Cancel {
-				return &BeforeCompactResult{
-					Cancel: true,
-					Reason: r.Reason,
+			if r, ok := result.(extensions.BeforeCompactResult); ok {
+				if r.Cancel {
+					return &BeforeCompactResult{
+						Cancel: true,
+						Reason: r.Reason,
+					}
+				}
+				if r.Summary != "" {
+					return &BeforeCompactResult{
+						Summary: r.Summary,
+					}
 				}
 			}
 			return nil

pkg/kit/hooks.go

@@ -107,12 +107,18 @@ type BeforeCompactHook struct {
 	IsAutomatic bool
 }
 
-// BeforeCompactResult controls whether compaction proceeds.
+// BeforeCompactResult controls whether compaction proceeds. Extensions can
+// cancel compaction or provide a custom summary that replaces the default
+// LLM-generated one.
 type BeforeCompactResult struct {
 	// Cancel, when true, prevents compaction from proceeding.
 	Cancel bool
 	// Reason is a human-readable explanation when Cancel is true.
 	Reason string
+	// Summary, when non-empty, replaces the default LLM-generated summary.
+	// The extension is responsible for generating a useful summary.
+	// Ignored when Cancel is true.
+	Summary string
 }
 
 // ---------------------------------------------------------------------------

pkg/kit/kit.go

@@ -62,6 +62,10 @@ type Kit struct {
 	// tool definitions, etc.
 	lastInputTokensMu sync.RWMutex
 	lastInputTokens   int
+
+	// subagentListeners holds per-tool-call event listeners registered via
+	// SubscribeSubagent(). Keyed by toolCallID → *subagentListenerSet.
+	subagentListeners sync.Map
 }
 
 // Subscribe registers an EventListener that will be called for every lifecycle
@@ -1401,17 +1405,23 @@ func (m *Kit) generate(ctx context.Context, messages []fantasy.Message) (*agent.
 	// spawn_subagent core tool can create child Kit instances without
 	// importing pkg/kit (which would create an import cycle).
 	ctx = core.WithSubagentSpawner(ctx, func(
-		spawnCtx context.Context, prompt, model, systemPrompt string, timeout time.Duration,
+		spawnCtx context.Context, toolCallID, prompt, model, systemPrompt string, timeout time.Duration,
 	) (*core.SubagentSpawnResult, error) {
+		// Build OnEvent: dispatch to per-tool-call listeners if any are
+		// registered via SubscribeSubagent(). Listeners are cleaned up
+		// after the subagent completes.
+		var onEvent func(Event)
+		if listeners := m.getSubagentListenerSet(toolCallID); listeners != nil {
+			onEvent = listeners.emit
+		}
 		result, err := m.Subagent(spawnCtx, SubagentConfig{
 			Prompt:       prompt,
 			Model:        model,
 			SystemPrompt: systemPrompt,
 			Timeout:      timeout,
-			OnEvent: func(e Event) {
-				m.events.emit(e)
-			},
+			OnEvent:      onEvent,
 		})
+		m.cleanupSubagentListeners(toolCallID)
 		if result == nil {
 			return &core.SubagentSpawnResult{Error: err}, err
 		}
@@ -1468,6 +1478,15 @@ func (m *Kit) generate(ctx context.Context, messages []fantasy.Message) (*agent.
 		func(delta string) {
 			m.events.emit(ReasoningDeltaEvent{Delta: delta})
 		},
+		func(toolCallID, toolName, chunk string, isStderr bool) {
+			// Emit tool output chunk event for streaming bash output
+			m.events.emit(ToolOutputEvent{
+				ToolCallID: toolCallID,
+				ToolName:   toolName,
+				Chunk:      chunk,
+				IsStderr:   isStderr,
+			})
+		},
 	)
 }
 
@@ -1554,6 +1573,18 @@ func (m *Kit) runTurn(ctx context.Context, promptLabel string, prompt string, pr
 
 	result, err := m.generate(ctx, messages)
 	if err != nil {
+		// Persist any messages from completed steps (tool call/result
+		// pairs) so partial progress is not lost. The agent layer only
+		// includes fully-paired tool_use + tool_result messages in
+		// completedStepMessages, so there are no orphaned entries that
+		// would break subsequent API requests. The user message and any
+		// completed work remain in the session; only the in-progress
+		// (pending) message or tool call is discarded.
+		if result != nil && len(result.ConversationMessages) > sentCount {
+			for _, msg := range result.ConversationMessages[sentCount:] {
+				_, _ = m.treeSession.AppendFantasyMessage(msg)
+			}
+		}
 		m.events.emit(TurnEndEvent{Error: err})
 		// Run AfterTurn hooks even on error.
 		if m.afterTurn.hasHooks() {
@@ -1752,6 +1783,30 @@ func (m *Kit) PromptResultWithFiles(ctx context.Context, message string, files [
 	})
 }
 
+// PromptResultWithMessages submits multiple user messages in a single turn.
+// All messages are persisted to the session and sent to the agent together.
+// The agent will respond once to the combined context of all messages.
+// Returns the full turn result including usage statistics and conversation messages.
+func (m *Kit) PromptResultWithMessages(ctx context.Context, messages []string) (*TurnResult, error) {
+	if len(messages) == 0 {
+		return nil, fmt.Errorf("no messages provided")
+	}
+
+	// Build prompt label from all messages
+	promptLabel := strings.Join(messages, " | ")
+	if len(promptLabel) > 100 {
+		promptLabel = promptLabel[:100] + "..."
+	}
+
+	// Build fantasy messages from all strings
+	var preMessages []fantasy.Message
+	for _, msg := range messages {
+		preMessages = append(preMessages, fantasy.NewUserMessage(msg))
+	}
+
+	return m.runTurn(ctx, promptLabel, messages[len(messages)-1], preMessages)
+}
+
 // ClearSession resets the tree session's leaf pointer to the root, starting
 // a fresh conversation branch.
 func (m *Kit) ClearSession() {

pkg/kit/sessions.go

@@ -34,6 +34,12 @@ func DeleteSession(path string) error {
 	return session.DeleteSession(path)
 }
 
+// OpenTreeSession opens an existing JSONL session file. This is a package-level
+// function (no Kit instance required) used by the CLI for session switching.
+func OpenTreeSession(path string) (*TreeManager, error) {
+	return session.OpenTreeSession(path)
+}
+
 // --- Instance methods on Kit ---
 
 // GetTreeSession returns the tree session manager, or nil if not configured.

skills/kit-sdk/SKILL.md

@@ -0,0 +1,772 @@
+---
+name: kit-sdk
+description: Guide for building Go applications with the Kit SDK. Use when the user asks to create a program, service, script, or application that uses Kit programmatically as a Go library — e.g. embedding LLM interactions, building agents, creating CLI tools powered by Kit, or integrating Kit into backend services. Do NOT use for Kit extensions (use kit-extensions skill instead).
+---
+
+# Kit SDK Development Guide
+
+The Kit SDK (`pkg/kit`) lets you embed Kit's full agent capabilities — LLM interactions, tool execution, session management, streaming, hooks — into any Go application. Unlike extensions (which are interpreted scripts running inside Kit's TUI), SDK programs are standalone compiled Go binaries.
+
+## Installation
+
+```bash
+go get github.com/mark3labs/kit
+```
+
+Import path (alias recommended):
+
+```go
+import kit "github.com/mark3labs/kit/pkg/kit"
+```
+
+## Quick Start
+
+```go
+package main
+
+import (
+    "context"
+    "fmt"
+    "log"
+
+    kit "github.com/mark3labs/kit/pkg/kit"
+)
+
+func main() {
+    ctx := context.Background()
+
+    host, err := kit.New(ctx, nil) // nil = load ~/.kit.yml defaults
+    if err != nil {
+        log.Fatal(err)
+    }
+    defer func() { _ = host.Close() }()
+
+    response, err := host.Prompt(ctx, "What is 2+2?")
+    if err != nil {
+        log.Fatal(err)
+    }
+    fmt.Println(response)
+}
+```
+
+## Core Lifecycle
+
+1. **Create**: `kit.New(ctx, opts)` — loads config, initializes MCP servers, creates LLM provider, sets up agent
+2. **Interact**: `host.Prompt(ctx, msg)` — send messages, agent uses tools as needed
+3. **Close**: `host.Close()` — cleans up MCP connections, model resources, session file handle
+
+Always defer `Close()`:
+
+```go
+defer func() { _ = host.Close() }()
+```
+
+---
+
+## Options Reference
+
+All fields are optional. Zero values use CLI defaults.
+
+```go
+host, err := kit.New(ctx, &kit.Options{
+    // Model
+    Model:        "anthropic/claude-sonnet-4-5-20250929", // "provider/model" format
+    SystemPrompt: "You are a helpful assistant",
+    ConfigFile:   "/path/to/config.yml",                  // default: ~/.kit.yml
+
+    // Behavior
+    MaxSteps:  10,   // 0 = unlimited tool-calling steps
+    Streaming: true, // stream LLM output (default from config)
+    Quiet:     true, // suppress debug output
+    Debug:     true, // enable debug logging
+
+    // Session
+    SessionDir:  "/path/to/project",  // base dir for session discovery (default: cwd)
+    SessionPath: "/path/to/session.jsonl", // open specific session file
+    Continue:    true,                // resume most recent session for SessionDir
+    NoSession:   true,                // ephemeral in-memory session, no disk persistence
+
+    // Tools
+    Tools:      []kit.Tool{kit.NewBashTool()}, // REPLACES entire default tool set
+    ExtraTools: []kit.Tool{myTool},            // ADDS alongside core/MCP/extension tools
+
+    // Skills
+    Skills:    []string{"/path/to/skill.md"}, // explicit skill files (empty = auto-discover)
+    SkillsDir: "/path/to/skills",             // override project-local skills dir
+
+    // Compaction
+    AutoCompact:       true,                        // auto-compact near context limit
+    CompactionOptions: &kit.CompactionOptions{...}, // nil = defaults
+})
+```
+
+**Critical distinction**: `Tools` replaces ALL default tools (core + MCP + extension). `ExtraTools` adds tools alongside the defaults. Use `Tools` to restrict the agent's capabilities; use `ExtraTools` to extend them.
+
+---
+
+## Prompt Methods
+
+### Simple prompt — string in, string out
+
+```go
+response, err := host.Prompt(ctx, "Explain this code")
+```
+
+### Full result with usage stats
+
+```go
+result, err := host.PromptResult(ctx, "Analyze this file")
+// result.Response     — assistant's text
+// result.StopReason   — "stop", "length", "tool-calls", "error", etc.
+// result.SessionID    — session UUID
+// result.TotalUsage   — aggregate tokens across all steps (*kit.FantasyUsage)
+// result.FinalUsage   — tokens from last API call only
+// result.Messages     — full updated conversation ([]kit.FantasyMessage)
+```
+
+### Multimodal with file attachments
+
+```go
+import "charm.land/fantasy"
+
+files := []fantasy.FilePart{{
+    Name:      "screenshot.png",
+    MediaType: "image/png",
+    Data:      imageBytes,
+}}
+result, err := host.PromptResultWithFiles(ctx, "What's in this image?", files)
+```
+
+### Per-call system message injection
+
+```go
+response, err := host.PromptWithOptions(ctx, "Review this PR", kit.PromptOptions{
+    SystemMessage: "Focus on security vulnerabilities only.",
+})
+```
+
+### System-level steering (no visible user message)
+
+```go
+response, err := host.Steer(ctx, "Switch to a more formal tone")
+```
+
+### Continue without new input
+
+```go
+response, err := host.FollowUp(ctx, "") // empty = "Continue."
+```
+
+### Multiple user messages in one turn
+
+```go
+result, err := host.PromptResultWithMessages(ctx, []string{
+    "Here is the code:",
+    "@file.go", // content from earlier
+    "Please review it.",
+})
+```
+
+### Legacy inline callbacks (deprecated — use event subscribers instead)
+
+```go
+response, err := host.PromptWithCallbacks(ctx, "List files",
+    func(name, args string) { fmt.Printf("Tool: %s\n", name) },
+    func(name, args, result string, isError bool) { /* tool result */ },
+    func(chunk string) { fmt.Print(chunk) }, // streaming
+)
+```
+
+---
+
+## Event System
+
+Events are read-only observations of the agent lifecycle. Register before calling Prompt.
+
+### Typed convenience subscribers
+
+```go
+// Each returns an unsubscribe function.
+unsub := host.OnToolCall(func(e kit.ToolCallEvent) {
+    // e.ToolCallID, e.ToolName, e.ToolKind, e.ToolArgs, e.ParsedArgs
+})
+defer unsub()
+
+host.OnToolResult(func(e kit.ToolResultEvent) {
+    // e.ToolCallID, e.ToolName, e.ToolKind, e.ToolArgs, e.ParsedArgs
+    // e.Result, e.IsError, e.Metadata (*ToolResultMetadata)
+})
+
+host.OnToolOutput(func(e kit.ToolOutputEvent) {
+    // e.ToolCallID, e.ToolName, e.Chunk, e.IsStderr
+    // Streaming bash output chunks
+})
+
+host.OnStreaming(func(e kit.MessageUpdateEvent) {
+    fmt.Print(e.Chunk) // real-time text streaming
+})
+
+host.OnResponse(func(e kit.ResponseEvent) {
+    // e.Content — final response text
+})
+
+host.OnTurnStart(func(e kit.TurnStartEvent) {
+    // e.Prompt
+})
+
+host.OnTurnEnd(func(e kit.TurnEndEvent) {
+    // e.Response, e.Error, e.StopReason
+})
+```
+
+### Generic subscriber (receives all events)
+
+```go
+unsub := host.Subscribe(func(e kit.Event) {
+    switch ev := e.(type) {
+    case kit.ToolCallEvent:
+        // ...
+    case kit.MessageUpdateEvent:
+        // ...
+    case kit.CompactionEvent:
+        // ev.Summary, ev.OriginalTokens, ev.CompactedTokens
+    }
+})
+```
+
+### All event types
+
+| Event Type | Struct | Key Fields |
+|------------|--------|------------|
+| `turn_start` | `TurnStartEvent` | `Prompt` |
+| `turn_end` | `TurnEndEvent` | `Response`, `Error`, `StopReason` |
+| `message_start` | `MessageStartEvent` | *(none)* |
+| `message_update` | `MessageUpdateEvent` | `Chunk` |
+| `message_end` | `MessageEndEvent` | `Content` |
+| `tool_call` | `ToolCallEvent` | `ToolCallID`, `ToolName`, `ToolKind`, `ToolArgs`, `ParsedArgs` |
+| `tool_execution_start` | `ToolExecutionStartEvent` | `ToolCallID`, `ToolName`, `ToolKind`, `ToolArgs` |
+| `tool_execution_end` | `ToolExecutionEndEvent` | `ToolCallID`, `ToolName`, `ToolKind` |
+| `tool_result` | `ToolResultEvent` | `ToolCallID`, `ToolName`, `ToolKind`, `ToolArgs`, `ParsedArgs`, `Result`, `IsError`, `Metadata` |
+| `tool_call_content` | `ToolCallContentEvent` | `Content` |
+| `tool_output` | `ToolOutputEvent` | `ToolCallID`, `ToolName`, `Chunk`, `IsStderr` |
+| `response` | `ResponseEvent` | `Content` |
+| `compaction` | `CompactionEvent` | `Summary`, `OriginalTokens`, `CompactedTokens`, `MessagesRemoved`, `ReadFiles`, `ModifiedFiles` |
+| `reasoning_delta` | `ReasoningDeltaEvent` | `Delta` |
+
+### Tool kind constants
+
+Tools are classified by kind for UI rendering:
+
+- `ToolKindExecute` = `"execute"` — bash
+- `ToolKindEdit` = `"edit"` — edit, write
+- `ToolKindRead` = `"read"` — read, ls
+- `ToolKindSearch` = `"search"` — grep, find
+- `ToolKindSubagent` = `"agent"` — spawn_subagent
+
+---
+
+## Hook System (Interceptors)
+
+Hooks can **modify or cancel** operations. Events are read-only; hooks are read-write.
+
+### BeforeToolCall — block tool execution
+
+```go
+unsub := host.OnBeforeToolCall(kit.HookPriorityNormal, func(h kit.BeforeToolCallHook) *kit.BeforeToolCallResult {
+    // h.ToolCallID, h.ToolName, h.ToolArgs
+    if h.ToolName == "bash" {
+        return &kit.BeforeToolCallResult{Block: true, Reason: "bash disabled"}
+    }
+    return nil // allow
+})
+```
+
+### AfterToolResult — modify tool output
+
+```go
+host.OnAfterToolResult(kit.HookPriorityNormal, func(h kit.AfterToolResultHook) *kit.AfterToolResultResult {
+    // h.ToolCallID, h.ToolName, h.ToolArgs, h.Result, h.IsError
+    if h.ToolName == "read" {
+        filtered := redactSecrets(h.Result)
+        return &kit.AfterToolResultResult{Result: &filtered}
+    }
+    return nil
+})
+```
+
+### BeforeTurn — modify prompt, inject messages
+
+```go
+host.OnBeforeTurn(kit.HookPriorityNormal, func(h kit.BeforeTurnHook) *kit.BeforeTurnResult {
+    // h.Prompt
+    newPrompt := h.Prompt + "\nAlways respond in JSON."
+    return &kit.BeforeTurnResult{Prompt: &newPrompt}
+    // Also available: SystemPrompt *string, InjectText *string
+})
+```
+
+### AfterTurn — observation only
+
+```go
+host.OnAfterTurn(kit.HookPriorityNormal, func(h kit.AfterTurnHook) {
+    // h.Response, h.Error
+    log.Printf("Turn completed: %d chars", len(h.Response))
+})
+```
+
+### ContextPrepare — filter/inject context window
+
+```go
+host.OnContextPrepare(kit.HookPriorityNormal, func(h kit.ContextPrepareHook) *kit.ContextPrepareResult {
+    // h.Messages — []fantasy.Message (the full context being sent to the LLM)
+    // Return nil to pass through, or replace entire context:
+    return &kit.ContextPrepareResult{Messages: filteredMessages}
+})
+```
+
+### BeforeCompact — cancel or customize compaction
+
+```go
+host.OnBeforeCompact(kit.HookPriorityNormal, func(h kit.BeforeCompactHook) *kit.BeforeCompactResult {
+    // h.EstimatedTokens, h.ContextLimit, h.UsagePercent, h.MessageCount, h.IsAutomatic
+    if h.IsAutomatic && h.UsagePercent < 0.9 {
+        return &kit.BeforeCompactResult{Cancel: true, Reason: "not yet"}
+    }
+    return nil
+})
+```
+
+### Hook priorities
+
+```go
+kit.HookPriorityHigh   = 0   // runs first
+kit.HookPriorityNormal = 50  // default
+kit.HookPriorityLow    = 100 // runs last
+```
+
+Lower values run first. Within the same priority, registration order applies. First non-nil result wins.
+
+---
+
+## Tools
+
+### Built-in tool constructors
+
+```go
+kit.NewReadTool(opts...)  // file reading
+kit.NewWriteTool(opts...) // file writing
+kit.NewEditTool(opts...)  // surgical text editing
+kit.NewBashTool(opts...)  // bash command execution
+kit.NewGrepTool(opts...) // content search (uses ripgrep when available)
+kit.NewFindTool(opts...) // file search (uses fd when available)
+kit.NewLsTool(opts...)   // directory listing
+```
+
+### Tool bundles
+
+```go
+kit.AllTools(opts...)       // all 7 core tools
+kit.CodingTools(opts...)    // bash, read, write, edit
+kit.ReadOnlyTools(opts...)  // read, grep, find, ls
+kit.SubagentTools(opts...)  // all except spawn_subagent (prevents recursion)
+```
+
+### Tool options
+
+```go
+kit.WithWorkDir("/path/to/dir") // override working directory for file-based tools
+```
+
+### Using tools in Options
+
+```go
+// Restricted: agent can ONLY run bash
+host, _ := kit.New(ctx, &kit.Options{
+    Tools: []kit.Tool{kit.NewBashTool()},
+})
+
+// Extended: all defaults PLUS a custom tool
+host, _ := kit.New(ctx, &kit.Options{
+    ExtraTools: []kit.Tool{myCustomTool},
+})
+```
+
+### Querying tools at runtime
+
+```go
+names := host.GetToolNames()       // []string of all tool names
+tools := host.GetTools()           // []kit.Tool (full tool objects)
+mcpCount := host.GetMCPToolCount() // tools from MCP servers
+extCount := host.GetExtensionToolCount() // tools from extensions
+```
+
+---
+
+## Session Management
+
+Sessions automatically persist as JSONL tree files. No explicit save needed.
+
+### Session modes (via Options)
+
+| Mode | Options | Behavior |
+|------|---------|----------|
+| Default | `{}` | New session file for cwd |
+| Specific file | `{SessionPath: "path.jsonl"}` | Open existing session |
+| Continue | `{Continue: true}` | Resume most recent session for cwd |
+| Ephemeral | `{NoSession: true}` | In-memory only, no disk persistence |
+| Custom dir | `{SessionDir: "/path"}` | Base directory for session discovery |
+
+### Instance methods
+
+```go
+host.GetSessionPath() // file path of active session
+host.GetSessionID()   // UUID of active session
+host.ClearSession()   // reset to fresh branch (doesn't delete file)
+host.Branch("entry-id") // branch from a specific entry
+host.SetSessionName("my session") // set display name
+
+// Get conversation messages
+msgs := host.GetSessionMessages()       // []extensions.SessionMessage (flattened text)
+msgs := host.GetStructuredMessages()     // []kit.StructuredMessage (typed content parts)
+```
+
+### Package-level session operations (no Kit instance needed)
+
+```go
+sessions, _ := kit.ListSessions("/path/to/project") // sessions for a directory
+sessions, _ := kit.ListAllSessions()                  // all sessions everywhere
+kit.DeleteSession("/path/to/session.jsonl")
+tm, _ := kit.OpenTreeSession("/path/to/session.jsonl") // open for direct access
+```
+
+---
+
+## Model Management
+
+### At creation time
+
+```go
+host, _ := kit.New(ctx, &kit.Options{
+    Model: "openai/gpt-4o",
+})
+```
+
+### At runtime
+
+```go
+err := host.SetModel(ctx, "anthropic/claude-sonnet-4-5-20250929")
+modelStr := host.GetModelString()   // "provider/model"
+info := host.GetModelInfo()          // *kit.ModelInfo (capabilities, limits, pricing) or nil
+isReasoning := host.IsReasoningModel()
+level := host.GetThinkingLevel()
+err = host.SetThinkingLevel(ctx, "medium") // recreates agent with new thinking budget
+```
+
+### Model registry
+
+```go
+models := host.GetAvailableModels()      // []extensions.ModelInfoEntry
+providers := kit.GetSupportedProviders() // []string
+providers := kit.GetFantasyProviders()   // providers usable with fantasy
+models, _ := kit.GetModelsForProvider("anthropic") // map[string]kit.ModelInfo
+info := kit.LookupModel("anthropic", "claude-sonnet-4-5-20250929") // *kit.ModelInfo
+info := kit.GetProviderInfo("openai")    // *kit.ProviderInfo (env vars, API URL)
+err := kit.ValidateEnvironment("anthropic", "") // check API keys
+suggestions := kit.SuggestModels("anthropic", "claudee") // fuzzy match
+kit.RefreshModelRegistry() // reload model database
+```
+
+### Model string format
+
+Always `"provider/model"`: `"anthropic/claude-sonnet-4-5-20250929"`, `"openai/gpt-4o"`, `"ollama/qwen3:8b"`.
+
+```go
+provider, modelID, err := kit.ParseModelString("anthropic/claude-sonnet-4-5-20250929")
+```
+
+---
+
+## Context & Compaction
+
+```go
+tokens := host.EstimateContextTokens()  // heuristic token count
+shouldCompact := host.ShouldCompact()    // true if near context limit
+
+stats := host.GetContextStats()
+// stats.EstimatedTokens — uses API-reported count when available (more accurate)
+// stats.ContextLimit    — model's context window size
+// stats.UsagePercent    — fraction used (0.0–1.0)
+// stats.MessageCount    — number of messages
+
+// Manual compaction
+result, err := host.Compact(ctx, nil, "") // nil opts = defaults, "" = default prompt
+// result.Summary, result.OriginalTokens, result.CompactedTokens, result.MessagesRemoved
+
+// Auto-compaction via Options
+host, _ := kit.New(ctx, &kit.Options{
+    AutoCompact: true,
+    CompactionOptions: &kit.CompactionOptions{
+        ReserveTokens:   16384,
+        KeepRecentTokens: 4096,
+        ContextWindow:   200000,
+    },
+})
+```
+
+---
+
+## In-Process Subagents
+
+Spawn child Kit instances without subprocess overhead:
+
+```go
+result, err := host.Subagent(ctx, kit.SubagentConfig{
+    Prompt:       "Analyze the test files and summarize coverage",
+    Model:        "anthropic/claude-haiku-3-5-20241022", // empty = parent's model
+    SystemPrompt: "You are a test analysis expert.",
+    Tools:        nil,           // nil = SubagentTools() (all except spawn_subagent)
+    NoSession:    true,          // ephemeral
+    Timeout:      2 * time.Minute, // 0 = 5 minute default
+    OnEvent: func(e kit.Event) {
+        // Real-time events from the child agent
+        if chunk, ok := e.(kit.MessageUpdateEvent); ok {
+            fmt.Print(chunk.Chunk)
+        }
+    },
+})
+// result.Response, result.Error, result.SessionID, result.StopReason
+// result.Usage (*kit.FantasyUsage), result.Elapsed (time.Duration)
+```
+
+### Subscribing to subagent events from parent
+
+```go
+host.OnToolCall(func(e kit.ToolCallEvent) {
+    if e.ToolName == "spawn_subagent" {
+        host.SubscribeSubagent(e.ToolCallID, func(child kit.Event) {
+            // Real-time events scoped to this subagent
+        })
+    }
+})
+```
+
+---
+
+## Authentication
+
+```go
+cm, _ := kit.NewCredentialManager()
+hasKey := kit.HasAnthropicCredentials()
+apiKey := kit.GetAnthropicAPIKey() // stored creds → ANTHROPIC_API_KEY env var
+```
+
+---
+
+## Skills
+
+```go
+// Load a single skill file
+skill, _ := kit.LoadSkill("/path/to/SKILL.md")
+// skill.Name, skill.Description, skill.Content, skill.Path
+
+// Load from directory
+skills, _ := kit.LoadSkillsFromDir("/path/to/skills")
+
+// Auto-discover (global + project-local)
+skills, _ := kit.LoadSkills("/path/to/project")
+
+// Prompt building with skills
+pb := kit.NewPromptBuilder("You are an assistant")
+pb.WithSkills(skills)
+pb.WithSection("", "Extra context here")
+systemPrompt := pb.Build()
+```
+
+---
+
+## Re-exported Types
+
+The SDK re-exports internal types so you don't need direct internal imports:
+
+```go
+// Message types
+kit.Message, kit.MessageRole, kit.ContentPart
+kit.TextContent, kit.ReasoningContent, kit.ToolCall, kit.ToolResult, kit.Finish
+kit.RoleUser, kit.RoleAssistant, kit.RoleTool, kit.RoleSystem
+
+// Session types
+kit.SessionInfo, kit.TreeManager, kit.SessionHeader, kit.MessageEntry
+
+// Config types
+kit.Config, kit.MCPServerConfig
+
+// Provider types
+kit.ProviderConfig, kit.ProviderResult, kit.ModelInfo, kit.ModelCost, kit.ModelLimit
+
+// Fantasy types (from charm.land/fantasy)
+kit.FantasyMessage, kit.FantasyUsage, kit.FantasyResponse
+
+// Compaction types
+kit.CompactionResult, kit.CompactionOptions
+
+// Conversion helpers
+msgs := kit.ConvertToFantasyMessages(&msg)   // SDK message → fantasy messages
+msg := kit.ConvertFromFantasyMessage(fMsg)    // fantasy message → SDK message
+```
+
+---
+
+## Common Patterns
+
+### Pattern: Scripting / CLI pipe
+
+Minimal program for automation — stdout-only output:
+
+```go
+host, _ := kit.New(ctx, &kit.Options{Quiet: true})
+defer func() { _ = host.Close() }()
+
+response, _ := host.Prompt(ctx, os.Args[1])
+fmt.Println(response)
+```
+
+### Pattern: Long-running autonomous agent
+
+Daemon that performs repeated independent tasks:
+
+```go
+host, _ := kit.New(ctx, &kit.Options{
+    SystemPrompt: taskPrompt,
+    Tools:        []kit.Tool{kit.NewBashTool()},
+    NoSession:    true,
+    Quiet:        true,
+})
+defer func() { _ = host.Close() }()
+
+ticker := time.NewTicker(30 * time.Minute)
+for {
+    select {
+    case <-ticker.C:
+        host.ClearSession() // fresh context each iteration
+        host.Prompt(ctx, "Perform the monitoring task")
+    case <-ctx.Done():
+        return
+    }
+}
+```
+
+### Pattern: Streaming output to terminal
+
+```go
+host.OnStreaming(func(e kit.MessageUpdateEvent) {
+    fmt.Print(e.Chunk)
+})
+response, _ := host.Prompt(ctx, "Write a poem")
+```
+
+### Pattern: Multi-turn conversation with memory
+
+```go
+host.Prompt(ctx, "My name is Alice")
+response, _ := host.Prompt(ctx, "What's my name?")
+// Session automatically maintains context across calls
+fmt.Printf("Session: %s\n", host.GetSessionPath())
+```
+
+### Pattern: Tool execution monitoring
+
+```go
+host.OnToolCall(func(e kit.ToolCallEvent) {
+    fmt.Printf("[%s] %s(%s)\n", e.ToolKind, e.ToolName, e.ToolArgs)
+})
+host.OnToolResult(func(e kit.ToolResultEvent) {
+    status := "✓"
+    if e.IsError { status = "✗" }
+    fmt.Printf("[%s] %s %s\n", e.ToolKind, status, e.ToolName)
+})
+```
+
+### Pattern: Guard rails with hooks
+
+```go
+// Block dangerous commands
+host.OnBeforeToolCall(kit.HookPriorityHigh, func(h kit.BeforeToolCallHook) *kit.BeforeToolCallResult {
+    if h.ToolName == "bash" && strings.Contains(h.ToolArgs, "rm -rf") {
+        return &kit.BeforeToolCallResult{Block: true, Reason: "dangerous command"}
+    }
+    return nil
+})
+
+// Inject context before every turn
+host.OnBeforeTurn(kit.HookPriorityNormal, func(h kit.BeforeTurnHook) *kit.BeforeTurnResult {
+    context := "Current user: admin\nEnvironment: production"
+    return &kit.BeforeTurnResult{InjectText: &context}
+})
+```
+
+### Pattern: Parallel subagents
+
+```go
+var wg sync.WaitGroup
+results := make([]*kit.SubagentResult, 3)
+
+tasks := []string{"Analyze auth module", "Analyze database layer", "Analyze API routes"}
+for i, task := range tasks {
+    wg.Add(1)
+    go func(idx int, t string) {
+        defer wg.Done()
+        results[idx], _ = host.Subagent(ctx, kit.SubagentConfig{
+            Prompt:    t,
+            NoSession: true,
+            Timeout:   3 * time.Minute,
+        })
+    }(i, task)
+}
+wg.Wait()
+```
+
+### Pattern: Read-only analysis agent
+
+```go
+host, _ := kit.New(ctx, &kit.Options{
+    SystemPrompt: "You are a code reviewer. Only read and analyze, never modify files.",
+    Tools:        kit.ReadOnlyTools(),
+})
+```
+
+---
+
+## Configuration
+
+The SDK loads config identically to the CLI:
+
+1. Explicit `ConfigFile` in Options (highest priority)
+2. `.kit.yml` in current directory
+3. `~/.kit.yml` in home directory
+4. Environment variables with `KIT_` prefix (`KIT_MODEL`, etc.)
+5. Provider-specific env vars (`ANTHROPIC_API_KEY`, `OPENAI_API_KEY`, etc.)
+
+Config files support `${ENV_VAR}` expansion.
+
+```go
+// Initialize config manually (usually not needed — kit.New handles this)
+kit.InitConfig("/path/to/config.yml", false)
+kit.LoadConfigWithEnvSubstitution("/path/to/config.yml")
+```
+
+---
+
+## Key Files for Reference
+
+- [`pkg/kit/kit.go`](https://github.com/mark3labs/kit/blob/main/pkg/kit/kit.go) — Kit struct, New(), Prompt methods, Subagent, Close
+- [`pkg/kit/types.go`](https://github.com/mark3labs/kit/blob/main/pkg/kit/types.go) — Re-exported types from internal packages
+- [`pkg/kit/tools.go`](https://github.com/mark3labs/kit/blob/main/pkg/kit/tools.go) — Tool constructors and bundles
+- [`pkg/kit/events.go`](https://github.com/mark3labs/kit/blob/main/pkg/kit/events.go) — Event types, EventBus, typed subscribers
+- [`pkg/kit/hooks.go`](https://github.com/mark3labs/kit/blob/main/pkg/kit/hooks.go) — Hook system (BeforeToolCall, AfterToolResult, etc.)
+- [`pkg/kit/sessions.go`](https://github.com/mark3labs/kit/blob/main/pkg/kit/sessions.go) — Session management
+- [`pkg/kit/compaction.go`](https://github.com/mark3labs/kit/blob/main/pkg/kit/compaction.go) — Context compaction
+- [`pkg/kit/models.go`](https://github.com/mark3labs/kit/blob/main/pkg/kit/models.go) — Model registry lookups
+- [`pkg/kit/config.go`](https://github.com/mark3labs/kit/blob/main/pkg/kit/config.go) — Config initialization and defaults
+- [`pkg/kit/skills.go`](https://github.com/mark3labs/kit/blob/main/pkg/kit/skills.go) — Skills loading and prompt building
+- [`pkg/kit/auth.go`](https://github.com/mark3labs/kit/blob/main/pkg/kit/auth.go) — Credential management
+- [`examples/sdk/`](https://github.com/mark3labs/kit/tree/main/examples/sdk) — Working example programs

www/pages/advanced/json-output.md

@@ -16,7 +16,7 @@ kit "Explain main.go" --json --quiet --no-session
 ```json
 {
   "response": "Final assistant response text",
-  "model": "anthropic/claude-haiku-3-5-20241022",
+  "model": "anthropic/claude-haiku-latest",
   "stop_reason": "end_turn",
   "session_id": "a1b2c3d4e5f6",
   "usage": {

www/pages/advanced/subagents.md

@@ -17,7 +17,7 @@ kit "Analyze codebase" \
     --no-session \
     --no-extensions \
     --quiet \
-    --model anthropic/claude-haiku-3-5-20241022
+    --model anthropic/claude-haiku-latest
 ```
 
 Key flags for subprocess usage:
@@ -39,7 +39,7 @@ Kit includes a built-in `spawn_subagent` tool that the LLM can use to delegate t
 ```
 spawn_subagent(
     task: "Analyze the test files and summarize coverage",
-    model: "anthropic/claude-haiku-3-5-20241022",   // optional
+    model: "anthropic/claude-haiku-latest",   // optional
     system_prompt: "You are a test analysis expert.",  // optional
     timeout_seconds: 300                               // optional, max 1800
 )
@@ -54,7 +54,7 @@ Extensions can spawn subagents programmatically:
 ```go
 result := ctx.SpawnSubagent(ext.SubagentConfig{
     Task:         "Review this code for security issues",
-    Model:        "anthropic/claude-sonnet-4-5-20250929",
+    Model:        "anthropic/claude-sonnet-latest",
     SystemPrompt: "You are a security auditor.",
 })
 ```
@@ -66,8 +66,33 @@ The SDK provides in-process subagent spawning:
 ```go
 result, err := host.Subagent(ctx, kit.SubagentConfig{
     Task:         "Summarize the changes in this PR",
-    Model:        "anthropic/claude-haiku-3-5-20241022",
+    Model:        "anthropic/claude-haiku-latest",
     SystemPrompt: "You are a code reviewer.",
     Timeout:      5 * time.Minute,
 })
 ```
+
+### Real-time subagent events
+
+Use `SubscribeSubagent` to receive real-time events from LLM-initiated subagents (i.e., when the model uses the `spawn_subagent` tool). Register inside an `OnToolCall` handler using the tool call ID:
+
+```go
+host.OnToolCall(func(e kit.ToolCallEvent) {
+    if e.ToolName == "spawn_subagent" {
+        host.SubscribeSubagent(e.ToolCallID, func(event kit.Event) {
+            switch ev := event.(type) {
+            case kit.MessageUpdateEvent:
+                fmt.Print(ev.Chunk) // streaming text from child
+            case kit.ToolCallEvent:
+                fmt.Printf("Child calling: %s\n", ev.ToolName)
+            case kit.ToolResultEvent:
+                fmt.Printf("Child result: %s\n", ev.ToolName)
+            }
+        })
+    }
+})
+```
+
+The listener receives the same event types as `Subscribe()` (`ToolCallEvent`, `MessageUpdateEvent`, `ReasoningDeltaEvent`, etc.) but scoped to the child agent's activity. Listeners are cleaned up automatically when the subagent completes.
+
+If no listeners are registered for a tool call, no event dispatching overhead is incurred.

www/pages/cli/commands.md

@@ -75,10 +75,71 @@ These commands are available inside the Kit TUI during an interactive session:
 | `/tree` | Navigate session tree |
 | `/fork` | Branch from an earlier message |
 | `/new` | Start a new session |
-| `/name` | Set session display name |
+| `/name [name]` | Set or show session display name |
+| `/resume` | Open session picker to switch sessions (alias: `/r`) |
 | `/session` | Show session info |
+| `/export [path]` | Export session as JSONL (default: auto-generated path) |
+| `/import <path>` | Import a session from a JSONL file |
+| `/share` | Upload session to GitHub Gist and get a shareable viewer URL |
 | `/quit` | Exit Kit |
 
+### Prompt history
+
+Use **↑** and **↓** arrow keys to navigate through previously submitted prompts. Kit keeps the last 100 entries. Consecutive duplicates are skipped.
+
+### Cancelling operations
+
+Press **ESC twice** to cancel the current operation:
+- During a tool call: rolls back the entire turn to maintain API message pairing
+- During streaming: stops the response generation
+
+This ensures that `tool_use` and `tool_result` messages are always sent to the API as matched pairs, avoiding errors from orphaned tool calls.
+
+## Prompt templates
+
+Create reusable prompt templates with shell-style argument substitution. Templates are loaded from `~/.kit/prompts/*.md` and `.kit/prompts/*.md`.
+
+### Creating templates
+
+Templates use YAML frontmatter for metadata and support argument placeholders:
+
+```markdown
+---
+description: Review code for issues
+---
+Review the following code for bugs and security issues.
+Focus on $1 specifically.
+```
+
+Save to `~/.kit/prompts/review.md` or `.kit/prompts/review.md`.
+
+### Using templates
+
+Templates appear as slash commands:
+
+```
+/review error handling
+```
+
+### Argument placeholders
+
+| Placeholder | Description |
+|-------------|-------------|
+| `$1`, `$2`, etc. | Individual arguments by position |
+| `$@`, `$ARGUMENTS` | All arguments joined with spaces |
+| `${@:N}` | Arguments from position N onwards |
+| `${@:N:L}` | L arguments starting at position N |
+
+### CLI flags
+
+```bash
+# Load a specific template by name
+kit --prompt-template review
+
+# Disable template loading
+kit --no-prompt-templates
+```
+
 ## ACP server
 
 Run Kit as an [ACP (Agent Client Protocol)](https://agentclientprotocol.com) agent server. ACP-compatible clients communicate with Kit over JSON-RPC 2.0 on stdin/stdout.

www/pages/cli/flags.md

@@ -11,7 +11,7 @@ All flags can be passed to the root `kit` command.
 
 | Flag | Short | Default | Description |
 |------|-------|---------|-------------|
-| `--model` | `-m` | `anthropic/claude-sonnet-4-5-20250929` | Model to use (provider/model format) |
+| `--model` | `-m` | `anthropic/claude-sonnet-latest` | Model to use (provider/model format) |
 | `--provider-api-key` | — | — | API key for the provider |
 | `--provider-url` | — | — | Base URL for provider API |
 | `--tls-skip-verify` | — | `false` | Skip TLS certificate verification |
@@ -45,6 +45,8 @@ These flags control Kit's behavior. When a prompt is passed as a positional argu
 |------|-------|---------|-------------|
 | `--extension` | `-e` | — | Load additional extension file(s) (repeatable) |
 | `--no-extensions` | — | `false` | Disable all extensions |
+| `--prompt-template` | — | — | Load a specific prompt template by name |
+| `--no-prompt-templates` | — | `false` | Disable prompt template loading |
 
 ## Generation parameters
 

www/pages/configuration.md

@@ -17,7 +17,7 @@ Kit looks for configuration in the following locations, in order of priority:
 Create `~/.kit.yml`:
 
 ```yaml
-model: anthropic/claude-sonnet-4-5-20250929
+model: anthropic/claude-sonnet-latest
 max-tokens: 4096
 temperature: 0.7
 stream: true
@@ -27,7 +27,7 @@ stream: true
 
 | Key | Type | Default | Description |
 |-----|------|---------|-------------|
-| `model` | string | `anthropic/claude-sonnet-4-5-20250929` | Model to use (provider/model format) |
+| `model` | string | `anthropic/claude-sonnet-latest` | Model to use (provider/model format) |
 | `max-tokens` | int | `4096` | Maximum tokens in response |
 | `temperature` | float | `0.7` | Randomness 0.0–1.0 |
 | `top-p` | float | `0.95` | Nucleus sampling 0.0–1.0 |
@@ -43,6 +43,8 @@ stream: true
 | `tls-skip-verify` | bool | `false` | Skip TLS certificate verification |
 | `stop-sequences` | list | — | Custom stop sequences |
 | `theme` | object or string | — | UI theme ([inline overrides or file path](/themes)) |
+| `prompt-templates` | bool | `true` | Enable prompt template loading |
+| `prompt-template` | string | — | Specific template to load by name |
 
 ## Environment variables
 
@@ -114,3 +116,17 @@ theme: "./themes/my-custom-theme.yml"
 ```
 
 See [Themes](/themes) for the full theme file format, built-in themes, and the extension theme API.
+
+## Preferences persistence
+
+Kit automatically saves your UI preferences across sessions to `~/.config/kit/preferences.yml`:
+
+- **Theme** — Set via `/theme <name>` or `ctx.SetTheme()`
+- **Model** — Set via `/model <name>` or the model selector
+- **Thinking level** — Set via `/thinking <level>` or Shift+Tab cycling
+
+These preferences are restored on next launch. Precedence (highest to lowest):
+1. CLI flags (`--model`, `--thinking-level`)
+2. Config file (`model:`, `thinking-level:`)
+3. Saved preferences (`~/.config/kit/preferences.yml`)
+4. Default values

www/pages/extensions/capabilities.md

@@ -7,7 +7,7 @@ description: All extension capabilities — lifecycle events, tools, commands, w
 
 ## Lifecycle events
 
-Extensions can hook into 18 lifecycle events:
+Extensions can hook into 20 lifecycle events:
 
 | Event | Description |
 |-------|-------------|
@@ -18,6 +18,7 @@ Extensions can hook into 18 lifecycle events:
 | `OnAgentEnd` | Agent loop completed |
 | `OnToolCall` | Tool call requested by the model |
 | `OnToolExecutionStart` | Tool execution beginning |
+| `OnToolOutput` | Streaming tool output chunk (for long-running tools) |
 | `OnToolExecutionEnd` | Tool execution completed |
 | `OnToolResult` | Tool result returned |
 | `OnInput` | User input received |
@@ -29,6 +30,7 @@ Extensions can hook into 18 lifecycle events:
 | `OnBeforeFork` | Before forking a conversation branch |
 | `OnBeforeSessionSwitch` | Before switching sessions |
 | `OnBeforeCompact` | Before conversation compaction |
+| `OnCustomEvent` | Custom inter-extension event received |
 
 ### Example
 
@@ -227,7 +229,7 @@ Spawn in-process child Kit instances:
 ```go
 result := ctx.SpawnSubagent(ext.SubagentConfig{
     Task:         "Analyze the test files and summarize coverage",
-    Model:        "anthropic/claude-haiku-3-5-20241022",
+    Model:        "anthropic/claude-haiku-latest",
     SystemPrompt: "You are a test analysis expert.",
 })
 ```

www/pages/extensions/examples.md

@@ -51,6 +51,12 @@ Kit ships with a rich set of example extensions in the `examples/extensions/` di
 | [`summarize.go`](https://github.com/mark3labs/kit/blob/master/examples/extensions/summarize.go) | Conversation summarization |
 | [`lsp-diagnostics.go`](https://github.com/mark3labs/kit/blob/master/examples/extensions/lsp-diagnostics.go) | LSP diagnostic integration |
 
+## Themes
+
+| Extension | Description |
+|-----------|-------------|
+| [`neon-theme.go`](https://github.com/mark3labs/kit/blob/master/examples/extensions/neon-theme.go) | Custom theme registration and switching |
+
 ## Multi-agent
 
 | Extension | Description |
@@ -74,3 +80,7 @@ Kit ships with a rich set of example extensions in the `examples/extensions/` di
 | [`kit-kit-agents/`](https://github.com/mark3labs/kit/tree/master/examples/extensions/kit-kit-agents) | Multi-agent orchestration example |
 | [`kit-telegram/`](https://github.com/mark3labs/kit/tree/master/examples/extensions/kit-telegram) | Telegram bot integration |
 | [`status-tools/`](https://github.com/mark3labs/kit/tree/master/examples/extensions/status-tools) | Status bar tool examples |
+
+## Project-local example
+
+The Kit repository also includes a project-local extension at `.kit/extensions/go-edit-lint.go` that demonstrates running `gopls` and `golangci-lint` on Go file edits. This serves as an example of how to create extensions specific to a project by placing them in the `.kit/extensions/` directory.

www/pages/providers.md

@@ -26,27 +26,55 @@ Kit supports a wide range of LLM providers through a unified `provider/model` st
 
 ```bash
 provider/model            # Standard format
-anthropic/claude-sonnet-4-5-20250929
+anthropic/claude-sonnet-latest
 openai/gpt-4o
 ollama/llama3
-google/gemini-2.0-flash-exp
+google/gemini-2.5-flash
 ```
 
 ## Model aliases
 
 Kit provides aliases for commonly used models:
 
+### Anthropic Claude
+
 ```bash
-claude-opus-latest        → claude-opus-4-20250514
-claude-sonnet-latest      → claude-sonnet-4-5-20250929
-claude-4-opus-latest      → claude-opus-4-20250514
-claude-4-sonnet-latest    → claude-sonnet-4-5-20250929
+claude-opus-latest        → claude-opus-4-6
+claude-sonnet-latest      → claude-sonnet-4-6
+claude-haiku-latest       → claude-haiku-4-5
+claude-4-opus-latest      → claude-opus-4-6
+claude-4-sonnet-latest    → claude-sonnet-4-6
+claude-4-haiku-latest     → claude-haiku-4-5
 claude-3-7-sonnet-latest  → claude-3-7-sonnet-20250219
 claude-3-5-sonnet-latest  → claude-3-5-sonnet-20241022
 claude-3-5-haiku-latest   → claude-3-5-haiku-20241022
 claude-3-opus-latest      → claude-3-opus-20240229
 ```
 
+### OpenAI GPT
+
+```bash
+o1-latest                 → o1
+o3-latest                 → o3
+o4-latest                 → o4-mini
+gpt-5-latest              → gpt-5.4
+gpt-5-chat-latest         → gpt-5.4
+gpt-4-latest              → gpt-4o
+gpt-4                     → gpt-4o
+gpt-3.5-latest            → gpt-3.5-turbo
+gpt-3.5                   → gpt-3.5-turbo
+codex-latest              → codex-mini-latest
+```
+
+### Google Gemini
+
+```bash
+gemini-pro-latest         → gemini-2.5-pro
+gemini-flash-latest       → gemini-2.5-flash
+gemini-flash              → gemini-2.5-flash
+gemini-pro                → gemini-2.5-pro
+```
+
 ## Specifying a model
 
 Via CLI flag:
@@ -59,7 +87,7 @@ kit -m ollama/llama3
 Via config file:
 
 ```yaml
-model: anthropic/claude-sonnet-4-5-20250929
+model: anthropic/claude-sonnet-latest
 ```
 
 Via environment variable:

www/pages/quick-start.md

@@ -28,7 +28,7 @@ kit @main.go @test.go "Review these files"
 Use a specific model:
 
 ```bash
-kit --model anthropic/claude-sonnet-4-5-20250929
+kit --model anthropic/claude-sonnet-latest
 ```
 
 ## Non-interactive mode

www/pages/sdk/callbacks.md

@@ -113,3 +113,25 @@ host.OnAfterTurn(0, func(ctx context.Context) error {
 ```
 
 The first argument is a priority (lower = runs first).
+
+## Subagent event monitoring
+
+Monitor real-time events from LLM-initiated subagents (when the model uses the `spawn_subagent` tool):
+
+```go
+host.OnToolCall(func(e kit.ToolCallEvent) {
+    if e.ToolName == "spawn_subagent" {
+        host.SubscribeSubagent(e.ToolCallID, func(event kit.Event) {
+            // Receives the same event types as Subscribe(), scoped to the child agent
+            switch ev := event.(type) {
+            case kit.MessageUpdateEvent:
+                fmt.Print(ev.Chunk)
+            case kit.ToolCallEvent:
+                fmt.Printf("Subagent calling: %s\n", ev.ToolName)
+            }
+        })
+    }
+})
+```
+
+`SubscribeSubagent` returns an unsubscribe function. Listeners are also cleaned up automatically when the subagent completes. See [Subagents](/advanced/subagents) for more details.

www/pages/sessions.md

@@ -19,6 +19,17 @@ Path separators in the working directory are replaced with `--`. For example, `/
 
 Each line in the session file is a JSON entry representing a message, tool call, model change, or extension data. The tree structure allows branching from any message to explore alternate paths.
 
+## Compaction
+
+When conversations grow long, Kit can compact them to free up context window space. The compaction system:
+
+- **Non-destructive**: Old messages remain on disk for history; only the LLM context is summarized
+- **File tracking**: Tracks which files were read and modified across compactions
+- **Split-turn handling**: Can summarize large single turns by splitting them
+- **Tool result truncation**: Caps tool output during serialization to stay within token budgets
+
+Use `/compact [focus]` to manually compact, or enable `--auto-compact` to compact automatically near the context limit.
+
 ## Resuming sessions
 
 ### Continue most recent
@@ -39,6 +50,8 @@ kit --resume
 kit -r
 ```
 
+The session picker supports search, scope/filter toggles (all sessions vs. current directory), and session deletion. You can also open it during a session with the `/resume` slash command.
+
 ### Open a specific session
 
 ```bash
@@ -46,6 +59,22 @@ kit --session path/to/session.jsonl
 kit -s path/to/session.jsonl
 ```
 
+## Session commands
+
+These slash commands are available during an interactive session:
+
+| Command | Description |
+|---------|-------------|
+| `/name [name]` | Set or display the session's display name |
+| `/session` | Show session info (path, ID, message count) |
+| `/resume` | Open the session picker to switch sessions |
+| `/export [path]` | Export session as JSONL (auto-generates path if omitted) |
+| `/import <path>` | Import and switch to a session from a JSONL file |
+| `/share` | Upload session to GitHub Gist and get a shareable viewer URL |
+| `/tree` | Navigate the session tree |
+| `/fork` | Branch from an earlier message |
+| `/new` | Start a fresh session |
+
 ## Ephemeral mode
 
 Run without creating a session file:
@@ -55,3 +84,25 @@ kit --no-session
 ```
 
 This is useful for one-off prompts, scripting, and subagent patterns where persistence isn't needed.
+
+## Sharing sessions
+
+The `/share` command uploads your session JSONL to GitHub Gist (via the `gh` CLI) and prints a shareable viewer URL:
+
+```
+/share
+```
+
+The viewer is available at `https://go-kit.dev/session/#GIST_ID` and supports all message types including text, reasoning blocks, tool calls, images, and model changes.
+
+You can also load any JSONL session via URL parameter: `https://go-kit.dev/session/?url=https://example.com/session.jsonl`
+
+## Preferences persistence
+
+Kit automatically saves your preferences across sessions to `~/.config/kit/preferences.yml`:
+
+- **Theme** — Set via `/theme <name>`
+- **Model** — Set via `/model <name>` or the model selector
+- **Thinking level** — Set via `/thinking <level>` or Shift+Tab cycling
+
+These preferences are restored on next launch. Precedence: CLI flag > config file > saved preference > default.

www/pages/themes.md

@@ -19,6 +19,8 @@ Switch themes at runtime with the `/theme` command:
 
 Run `/theme` with no arguments to list all available themes.
 
+**Theme selections are automatically saved** to `~/.config/kit/preferences.yml` and restored on next launch. You don't need to add anything to your config file — just `/theme <name>` and it sticks. This persistence also applies to **model** and **thinking level** selections.
+
 ## Built-in themes
 
 | Theme | Style |
@@ -276,4 +278,14 @@ When multiple sources define the same theme name, later sources win:
 3. Project-local themes (`.kit/themes/`)
 4. Extension-registered themes (highest)
 
+### Startup theme resolution
+
+At startup, Kit determines which theme to apply:
+
+1. **`.kit.yml` `theme:` key** — explicit config always wins (highest priority)
+2. **`~/.config/kit/preferences.yml`** — persisted `/theme` selection
+3. **Default `kitt` theme** — fallback
+
+The preferences file is updated automatically whenever you use `/theme` or `ctx.SetTheme()`. It is separate from `.kit.yml` so it never clobbers your config comments or formatting.
+
 Theme changes via `/theme` or `ctx.SetTheme()` take effect immediately on all UI elements, including previously rendered messages.

www/tome.config.js

@@ -34,6 +34,7 @@ export default {
         "extensions/capabilities",
         "extensions/examples",
         "extensions/loading",
+        "extensions/testing",
       ],
     },
     {