chore: move go-edit-lint extension to global scope

- Remove .kit/extensions/go-edit-lint.go from the repo since the
  extension is now installed under ~/.config/kit/extensions/ for
  per-user use across all projects.

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

@@ -1,268 +0,0 @@
-//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
-}
-
-// Package-level state: set of .go files edited during the current agent turn.
-var editedFiles map[string]bool
-
-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 after agent turns that edit Go files")
-	})
-
-	// Track edited .go files — don't lint yet.
-	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
-		}
-
-		if editedFiles == nil {
-			editedFiles = make(map[string]bool)
-		}
-		editedFiles[absPath] = true
-		return nil
-	})
-
-	// After the agent turn ends, lint all collected files.
-	api.OnAgentEnd(func(e ext.AgentEndEvent, ctx ext.Context) {
-		if len(editedFiles) == 0 {
-			return
-		}
-
-		// Snapshot and reset immediately so the next turn starts clean.
-		files := editedFiles
-		editedFiles = nil
-
-		// Skip lint on errored turns.
-		if e.StopReason == "error" {
-			return
-		}
-
-		// Collect unique directories and file list for gopls.
-		var allGoplsOutput []string
-		for absPath := range files {
-			res := runGopls(ctx.CWD, absPath)
-			formatted := formatToolResult(res, "")
-			if formatted != "" {
-				allGoplsOutput = append(allGoplsOutput, fmt.Sprintf("# %s\n%s", filepath.Base(absPath), formatted))
-			}
-		}
-
-		lintRes := runGolangCILint(ctx.CWD, "./...")
-
-		goplsSection := "No diagnostics."
-		if len(allGoplsOutput) > 0 {
-			goplsSection = strings.Join(allGoplsOutput, "\n\n")
-		}
-		lintSection := formatToolResult(lintRes, "No lint issues.")
-
-		// Build file list for the report header.
-		var fileNames []string
-		for absPath := range files {
-			fileNames = append(fileNames, filepath.Base(absPath))
-		}
-
-		report := fmt.Sprintf(
-			"<go_diagnostics files=%q>\n[gopls]\n%s\n\n[golangci-lint]\n%s\n</go_diagnostics>",
-			strings.Join(fileNames, ", "),
-			goplsSection,
-			lintSection,
-		)
-
-		goplsIssues, lintIssues := countIssues(report)
-		hasIssues := goplsIssues > 0 || lintIssues > 0
-
-		if hasIssues {
-			// Show TUI block so the user sees it too.
-			var msgLines []string
-			msgLines = append(msgLines, fmt.Sprintf("Files: %s", strings.Join(fileNames, ", ")))
-			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))
-			}
-
-			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",
-			})
-
-			// Inject a follow-up message so the agent fixes the issues.
-			ctx.SendMessage(report + "\n\n⚠️ DIAGNOSTICS FOUND: Please review and fix the issues above.")
-		} else {
-			ctx.PrintBlock(ext.PrintBlockOpts{
-				Text:        fmt.Sprintf("Files: %s\n✓ All clean", strings.Join(fileNames, ", ")),
-				BorderColor: "#a6e3a1",
-				Subtitle:    "go-edit-lint",
-			})
-		}
-	})
-}
-
-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 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 {
-			if emptyFallback != "" {
-				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) {
-	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]
-		for _, line := range strings.Split(goplsSection, "\n") {
-			line = strings.TrimSpace(line)
-			if line != "" && line != "[gopls]" && line != "No diagnostics." && !strings.HasPrefix(line, "#") {
-				goplsCount++
-			}
-		}
-	}
-
-	if lintStart != -1 && endTag != -1 {
-		lintSection := report[lintStart:endTag]
-		for _, line := range strings.Split(lintSection, "\n") {
-			line = strings.TrimSpace(line)
-			if line != "" && line != "[golangci-lint]" && line != "No lint issues." {
-				lintCount++
-			}
-		}
-	}
-
-	return goplsCount, lintCount
-}

.kit/extensions/subagent-monitor.go

@@ -1,304 +0,0 @@
-//go:build ignore
-
-// subagent-monitor — live horizontal widget strip for spawned subagents
-//
-// Subscribes to subagents spawned by the main Kit agent and displays a
-// single widget just above the input box. Each subagent occupies one column
-// in a side-by-side horizontal layout. Columns show scrolling real-time
-// output as the subagent works. When a subagent finishes its column is
-// removed automatically.
-//
-// Yaegi-safe design notes:
-// - No sync.Mutex (Yaegi has reflection issues with sync primitives)
-// - No channels in maps (Yaegi panics on range over map[string]chan)
-// - All ctx.* calls guarded with nil checks
-// - Simple data structures only
-package main
-
-import (
-	"fmt"
-	"strings"
-	"time"
-
-	"kit/ext"
-)
-
-// ---------------------------------------------------------------------------
-// Per-subagent state
-// ---------------------------------------------------------------------------
-
-type submonEntry struct {
-	id      int
-	callID  string
-	task    string
-	lines   []string
-	started time.Time
-	elapsed time.Duration
-}
-
-const (
-	submonColWidth = 34 // visible character width per column
-	submonMaxLines = 5  // scrolling output lines per column
-	submonColGap   = 2  // spaces between columns
-)
-
-// ---------------------------------------------------------------------------
-// Package-level state - all simple types
-// ---------------------------------------------------------------------------
-
-var (
-	submonCtx     ext.Context
-	submonHasCtx  bool
-	submonEntries []*submonEntry
-	submonNextID  int
-)
-
-func submonInit() {
-	submonEntries = nil
-	submonNextID = 1
-}
-
-// ---------------------------------------------------------------------------
-// String helpers
-// ---------------------------------------------------------------------------
-
-func submonPad(s string, w int) string {
-	r := []rune(s)
-	if len(r) >= w {
-		return string(r[:w])
-	}
-	return s + strings.Repeat(" ", w-len(r))
-}
-
-func submonTrunc(s string, w int) string {
-	r := []rune(s)
-	if len(r) <= w {
-		return s
-	}
-	if w <= 1 {
-		return "…"
-	}
-	return string(r[:w-1]) + "…"
-}
-
-// ---------------------------------------------------------------------------
-// Widget rendering
-// ---------------------------------------------------------------------------
-
-func submonRenderColumn(e *submonEntry) []string {
-	var rows []string
-
-	// Calculate elapsed time on-demand to avoid race conditions with ticker
-	elapsed := e.elapsed
-	if elapsed == 0 && !e.started.IsZero() {
-		elapsed = time.Since(e.started)
-	}
-	secs := int(elapsed.Seconds())
-	timeStr := fmt.Sprintf("%ds", secs)
-	taskMax := submonColWidth - len(timeStr) - 3
-	taskPart := submonTrunc(e.task, taskMax)
-	header := fmt.Sprintf("#%d %s  %s", e.id, taskPart, timeStr)
-	rows = append(rows, submonPad(header, submonColWidth))
-
-	display := e.lines
-	if len(display) > submonMaxLines {
-		display = display[len(display)-submonMaxLines:]
-	}
-	for _, l := range display {
-		rows = append(rows, submonPad("  "+submonTrunc(l, submonColWidth-2), submonColWidth))
-	}
-	for len(rows) < submonMaxLines+1 {
-		if len(rows) == 1 && len(e.lines) == 0 {
-			rows = append(rows, submonPad("  waiting…", submonColWidth))
-		} else {
-			rows = append(rows, strings.Repeat(" ", submonColWidth))
-		}
-	}
-	return rows
-}
-
-func submonBuildWidget() string {
-	if len(submonEntries) == 0 {
-		return ""
-	}
-
-	numCols := len(submonEntries)
-	numRows := submonMaxLines + 1
-	cols := make([][]string, numCols)
-	for i, e := range submonEntries {
-		rows := submonRenderColumn(e)
-		col := make([]string, numRows)
-		for j := 0; j < numRows; j++ {
-			if j < len(rows) {
-				col[j] = rows[j]
-			} else {
-				col[j] = strings.Repeat(" ", submonColWidth)
-			}
-		}
-		cols[i] = col
-	}
-
-	gap := strings.Repeat(" ", submonColGap)
-	var sb strings.Builder
-	for row := 0; row < numRows; row++ {
-		for ci := range cols {
-			if ci > 0 {
-				sb.WriteString(gap)
-			}
-			sb.WriteString(cols[ci][row])
-		}
-		if row < numRows-1 {
-			sb.WriteString("\n")
-		}
-	}
-	return sb.String()
-}
-
-func submonPushWidget() {
-	if !submonHasCtx {
-		return
-	}
-	if submonCtx.SetWidget == nil {
-		return
-	}
-
-	text := submonBuildWidget()
-	if len(submonEntries) == 0 {
-		if submonCtx.RemoveWidget != nil {
-			submonCtx.RemoveWidget("submon")
-		}
-		return
-	}
-	submonCtx.SetWidget(ext.WidgetConfig{
-		ID:        "submon",
-		Placement: ext.WidgetAbove,
-		Content:   ext.WidgetContent{Text: text},
-		Style:     ext.WidgetStyle{BorderColor: "#89b4fa"},
-		Priority:  0,
-	})
-}
-
-func submonAppendLine(e *submonEntry, line string) {
-	line = strings.TrimRight(line, "\r\n")
-	if strings.TrimSpace(line) == "" {
-		return
-	}
-	e.lines = append(e.lines, line)
-}
-
-// ---------------------------------------------------------------------------
-// Init
-// ---------------------------------------------------------------------------
-
-func Init(api ext.API) {
-	submonInit()
-
-	api.OnSessionStart(func(_ ext.SessionStartEvent, ctx ext.Context) {
-		submonCtx = ctx
-		submonHasCtx = true
-		submonInit()
-		if ctx.RemoveWidget != nil {
-			ctx.RemoveWidget("submon")
-		}
-	})
-
-	api.OnAgentEnd(func(_ ext.AgentEndEvent, ctx ext.Context) {
-		submonCtx = ctx
-		submonHasCtx = true
-	})
-
-	// ── SubagentStart ────────────────────────────────────────────────────────
-	api.OnSubagentStart(func(e ext.SubagentStartEvent, ctx ext.Context) {
-		submonCtx = ctx
-		submonHasCtx = true
-
-		id := submonNextID
-		submonNextID++
-		entry := &submonEntry{
-			id:      id,
-			callID:  e.ToolCallID,
-			task:    e.Task,
-			started: time.Now(),
-		}
-		submonEntries = append(submonEntries, entry)
-
-		submonPushWidget()
-	})
-
-	// ── SubagentChunk ────────────────────────────────────────────────────────
-	api.OnSubagentChunk(func(e ext.SubagentChunkEvent, ctx ext.Context) {
-		submonCtx = ctx
-		submonHasCtx = true
-
-		var entry *submonEntry
-		for _, en := range submonEntries {
-			if en.callID == e.ToolCallID {
-				entry = en
-				break
-			}
-		}
-		if entry == nil {
-			return
-		}
-
-		switch e.ChunkType {
-		case "text":
-			for _, line := range strings.Split(e.Content, "\n") {
-				submonAppendLine(entry, line)
-			}
-		case "tool_call":
-			submonAppendLine(entry, "→ "+e.ToolName)
-		case "tool_execution_start":
-			submonAppendLine(entry, "⚙ "+e.ToolName)
-		case "tool_result":
-			if e.IsError {
-				submonAppendLine(entry, "✗ "+e.ToolName)
-			} else {
-				submonAppendLine(entry, "✓ "+e.ToolName)
-			}
-		}
-
-		submonPushWidget()
-	})
-
-	// ── SubagentEnd ──────────────────────────────────────────────────────────
-	api.OnSubagentEnd(func(e ext.SubagentEndEvent, ctx ext.Context) {
-		submonCtx = ctx
-		submonHasCtx = true
-
-		var entry *submonEntry
-		for _, en := range submonEntries {
-			if en.callID == e.ToolCallID {
-				entry = en
-				break
-			}
-		}
-		if entry != nil {
-			entry.elapsed = time.Since(entry.started)
-			if e.ErrorMsg != "" {
-				submonAppendLine(entry, "✗ "+submonTrunc(e.ErrorMsg, submonColWidth-2))
-			}
-		}
-
-		submonPushWidget()
-
-		// Remove the entry immediately (no goroutine to avoid races)
-		newEntries := submonEntries[:0]
-		for _, en := range submonEntries {
-			if en.callID != e.ToolCallID {
-				newEntries = append(newEntries, en)
-			}
-		}
-		submonEntries = newEntries
-		submonPushWidget()
-	})
-
-	// ── SessionShutdown ──────────────────────────────────────────────────────
-	api.OnSessionShutdown(func(_ ext.SessionShutdownEvent, ctx ext.Context) {
-		submonInit()
-		// Guard ctx access - may be nil during shutdown
-		if ctx.RemoveWidget != nil {
-			ctx.RemoveWidget("submon")
-		}
-	})
-}

cmd/root.go

@@ -434,6 +434,36 @@ func extensionCommandsForUI(k *kit.Kit) []commands.ExtensionCommand {
 	return cmds
 }
 
+// buildExtensionItems converts the loaded extensions into ui.ExtensionItem
+// values for the startup [Extensions] section. The display name is the file
+// basename without the .go suffix; for subdirectory extensions the parent
+// directory name is used (matching `kit extensions list`).
+func buildExtensionItems(k *kit.Kit, cwd string) []ui.ExtensionItem {
+	infos := k.Extensions().Loaded()
+	if len(infos) == 0 {
+		return nil
+	}
+	items := make([]ui.ExtensionItem, 0, len(infos))
+	for _, info := range infos {
+		name := filepath.Base(info.Path)
+		if name == "main.go" {
+			// Subdirectory extension: use the parent directory's name.
+			name = filepath.Base(filepath.Dir(info.Path))
+		}
+		name = strings.TrimSuffix(name, ".go")
+		source := "user"
+		if cwd != "" && strings.HasPrefix(info.Path, cwd) {
+			source = "project"
+		}
+		items = append(items, ui.ExtensionItem{
+			Name:   name,
+			Path:   info.Path,
+			Source: source,
+		})
+	}
+	return items
+}
+
 // widgetProviderForUI returns a function that converts extension widgets to
 // ui.WidgetData for the given placement. Returns nil if extensions are
 // disabled, which is safe — the UI treats a nil GetWidgets as "no widgets".
@@ -935,12 +965,17 @@ func runNormalMode(ctx context.Context) error {
 			source = "project"
 		}
 		skillItems = append(skillItems, ui.SkillItem{
-			Name:   s.Name,
-			Path:   s.Path,
-			Source: source,
+			Name:        s.Name,
+			Path:        s.Path,
+			Source:      source,
+			Description: s.Description,
 		})
 	}
 
+	// Build extension items from the loaded extensions for the [Extensions]
+	// startup section. Each entry is a single .go file (or a subdir's main.go).
+	extensionItems := buildExtensionItems(kitInstance, cwd)
+
 	// Build prompt template and skill item provider callbacks for hot-reload.
 	// These are called by the TUI when ContentReloadEvent fires.
 	getPromptTemplates := func() []*prompts.PromptTemplate {
@@ -976,14 +1011,22 @@ func runNormalMode(ctx context.Context) error {
 				source = "project"
 			}
 			items = append(items, ui.SkillItem{
-				Name:   s.Name,
-				Path:   s.Path,
-				Source: source,
+				Name:        s.Name,
+				Path:        s.Path,
+				Source:      source,
+				Description: s.Description,
 			})
 		}
 		return items
 	}
 
+	// getExtensionItems re-collects the loaded extension list, used by the
+	// TUI after an extension hot-reload to refresh the [Extensions] row.
+	getExtensionItems := func() []ui.ExtensionItem {
+		cwd, _ := os.Getwd()
+		return buildExtensionItems(kitInstance, cwd)
+	}
+
 	// Build extension UI providers once (shared between both modes).
 	getWidgets := widgetProviderForUI(kitInstance)
 	getHeader := headerProviderForUI(kitInstance)
@@ -1220,7 +1263,7 @@ func runNormalMode(ctx context.Context) error {
 
 	// 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, promptTemplates, contextPaths, skillItems, getPromptTemplates, getSkillItems, getToolNames, getMCPToolCount, mcpPrompts, getMCPPrompts, expandMCPPrompt, getWidgets, getHeader, getFooter, getToolRenderer, getEditorInterceptor, getUIVisibility, getStatusBarEntries, emitBeforeFork, emitBeforeSessionSwitch, getGlobalShortcuts, getExtensionCommands, setModelForUI, emitModelChangeForUI, kitInstance.IsReasoningModel(), kitInstance.GetThinkingLevel(), setThinkingLevelForUI, switchSessionForUI, reloadExtensionsForUI)
+		return runNonInteractiveModeApp(ctx, appInstance, cli, positionalPrompt, quietFlag, jsonFlag, noExitFlag, modelName, parsedProvider, kitInstance.GetLoadingMessage(), serverNames, toolNames, mcpToolCount, extensionToolCount, usageTracker, extCommands, promptTemplates, contextPaths, skillItems, extensionItems, getPromptTemplates, getSkillItems, getExtensionItems, getToolNames, getMCPToolCount, mcpPrompts, getMCPPrompts, expandMCPPrompt, getWidgets, getHeader, getFooter, getToolRenderer, getEditorInterceptor, getUIVisibility, getStatusBarEntries, emitBeforeFork, emitBeforeSessionSwitch, getGlobalShortcuts, getExtensionCommands, setModelForUI, emitModelChangeForUI, kitInstance.IsReasoningModel(), kitInstance.GetThinkingLevel(), setThinkingLevelForUI, switchSessionForUI, reloadExtensionsForUI)
 	}
 
 	// Quiet mode is not allowed in interactive mode
@@ -1228,7 +1271,7 @@ 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, promptTemplates, contextPaths, skillItems, getPromptTemplates, getSkillItems, getToolNames, getMCPToolCount, mcpPrompts, getMCPPrompts, expandMCPPrompt, getWidgets, getHeader, getFooter, getToolRenderer, getEditorInterceptor, getUIVisibility, getStatusBarEntries, emitBeforeFork, emitBeforeSessionSwitch, getGlobalShortcuts, getExtensionCommands, setModelForUI, emitModelChangeForUI, kitInstance.IsReasoningModel(), kitInstance.GetThinkingLevel(), setThinkingLevelForUI, switchSessionForUI, reloadExtensionsForUI, startupExtensionMessages)
+	return runInteractiveModeBubbleTea(ctx, appInstance, modelName, parsedProvider, kitInstance.GetLoadingMessage(), serverNames, toolNames, mcpToolCount, extensionToolCount, usageTracker, extCommands, promptTemplates, contextPaths, skillItems, extensionItems, getPromptTemplates, getSkillItems, getExtensionItems, getToolNames, getMCPToolCount, mcpPrompts, getMCPPrompts, expandMCPPrompt, getWidgets, getHeader, getFooter, getToolRenderer, getEditorInterceptor, getUIVisibility, getStatusBarEntries, emitBeforeFork, emitBeforeSessionSwitch, getGlobalShortcuts, getExtensionCommands, setModelForUI, emitModelChangeForUI, kitInstance.IsReasoningModel(), kitInstance.GetThinkingLevel(), setThinkingLevelForUI, switchSessionForUI, reloadExtensionsForUI, startupExtensionMessages)
 }
 
 // runNonInteractiveModeApp executes a single prompt via the app layer and exits,
@@ -1241,7 +1284,7 @@ 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 []commands.ExtensionCommand, promptTemplates []*prompts.PromptTemplate, contextPaths []string, skillItems []ui.SkillItem, getPromptTemplates func() []*prompts.PromptTemplate, getSkillItems func() []ui.SkillItem, getToolNames func() []string, getMCPToolCount func() int, mcpPrompts []ui.MCPPromptInfo, getMCPPrompts func() []ui.MCPPromptInfo, expandMCPPrompt func(string, string, map[string]string) (*ui.MCPPromptExpandResult, error), 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() []commands.ExtensionCommand, setModel func(string) error, emitModelChange func(string, string, string), isReasoningModel bool, thinkingLevel string, setThinkingLevel func(string) error, switchSession func(string) error, reloadExtensions func() 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 []commands.ExtensionCommand, promptTemplates []*prompts.PromptTemplate, contextPaths []string, skillItems []ui.SkillItem, extensionItems []ui.ExtensionItem, getPromptTemplates func() []*prompts.PromptTemplate, getSkillItems func() []ui.SkillItem, getExtensionItems func() []ui.ExtensionItem, getToolNames func() []string, getMCPToolCount func() int, mcpPrompts []ui.MCPPromptInfo, getMCPPrompts func() []ui.MCPPromptInfo, expandMCPPrompt func(string, string, map[string]string) (*ui.MCPPromptExpandResult, error), 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() []commands.ExtensionCommand, setModel func(string) error, emitModelChange func(string, string, string), isReasoningModel bool, thinkingLevel string, setThinkingLevel func(string) error, switchSession func(string) error, reloadExtensions func() error) error {
 	// Expand @file references in the prompt before sending to the agent.
 	// Text files are XML-inlined; binary files are extracted as multimodal parts.
 	var fileParts []kit.LLMFilePart
@@ -1302,7 +1345,7 @@ 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, promptTemplates, contextPaths, skillItems, getPromptTemplates, getSkillItems, getToolNames, getMCPToolCount, mcpPrompts, getMCPPrompts, expandMCPPrompt, getWidgets, getHeader, getFooter, getToolRenderer, getEditorInterceptor, getUIVisibility, getStatusBarEntries, emitBeforeFork, emitBeforeSessionSwitch, getGlobalShortcuts, getExtensionCommands, setModel, emitModelChange, isReasoningModel, thinkingLevel, setThinkingLevel, switchSession, reloadExtensions, nil)
+		return runInteractiveModeBubbleTea(ctx, appInstance, modelName, providerName, loadingMessage, serverNames, toolNames, mcpToolCount, extensionToolCount, usageTracker, extCommands, promptTemplates, contextPaths, skillItems, extensionItems, getPromptTemplates, getSkillItems, getExtensionItems, getToolNames, getMCPToolCount, mcpPrompts, getMCPPrompts, expandMCPPrompt, getWidgets, getHeader, getFooter, getToolRenderer, getEditorInterceptor, getUIVisibility, getStatusBarEntries, emitBeforeFork, emitBeforeSessionSwitch, getGlobalShortcuts, getExtensionCommands, setModel, emitModelChange, isReasoningModel, thinkingLevel, setThinkingLevel, switchSession, reloadExtensions, nil)
 	}
 
 	return nil
@@ -1400,7 +1443,7 @@ 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 []commands.ExtensionCommand, promptTemplates []*prompts.PromptTemplate, contextPaths []string, skillItems []ui.SkillItem, getPromptTemplates func() []*prompts.PromptTemplate, getSkillItems func() []ui.SkillItem, getToolNames func() []string, getMCPToolCount func() int, mcpPrompts []ui.MCPPromptInfo, getMCPPrompts func() []ui.MCPPromptInfo, expandMCPPrompt func(string, string, map[string]string) (*ui.MCPPromptExpandResult, error), 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() []commands.ExtensionCommand, setModel func(string) error, emitModelChange func(string, string, string), isReasoningModel bool, thinkingLevel string, setThinkingLevel func(string) error, switchSession func(string) error, reloadExtensions func() error, startupExtensionMessages []string) error {
+func runInteractiveModeBubbleTea(_ context.Context, appInstance *app.App, modelName, providerName, loadingMessage string, serverNames, toolNames []string, mcpToolCount, extensionToolCount int, usageTracker *ui.UsageTracker, extCommands []commands.ExtensionCommand, promptTemplates []*prompts.PromptTemplate, contextPaths []string, skillItems []ui.SkillItem, extensionItems []ui.ExtensionItem, getPromptTemplates func() []*prompts.PromptTemplate, getSkillItems func() []ui.SkillItem, getExtensionItems func() []ui.ExtensionItem, getToolNames func() []string, getMCPToolCount func() int, mcpPrompts []ui.MCPPromptInfo, getMCPPrompts func() []ui.MCPPromptInfo, expandMCPPrompt func(string, string, map[string]string) (*ui.MCPPromptExpandResult, error), 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() []commands.ExtensionCommand, setModel func(string) error, emitModelChange func(string, string, string), isReasoningModel bool, thinkingLevel string, setThinkingLevel func(string) error, switchSession func(string) error, reloadExtensions func() error, startupExtensionMessages []string) error {
 	// Redirect all log output (stdlib and charm) to a file so that log
 	// messages don't write to stderr and corrupt the TUI. Bubble Tea
 	// captures stdout for rendering; any stray stderr output from
@@ -1445,6 +1488,8 @@ func runInteractiveModeBubbleTea(_ context.Context, appInstance *app.App, modelN
 		ContextPaths:             contextPaths,
 		SkillItems:               skillItems,
 		GetSkillItems:            getSkillItems,
+		ExtensionItems:           extensionItems,
+		GetExtensionItems:        getExtensionItems,
 		StartupExtensionMessages: startupExtensionMessages,
 		GetWidgets:               getWidgets,
 		GetHeader:                getHeader,

internal/agent/mcp_adapter.go

@@ -9,12 +9,19 @@ import (
 	"github.com/mark3labs/kit/internal/tools"
 )
 
+// mcpExecutor is the subset of *tools.MCPToolManager that the adapter
+// actually uses. Extracted as an interface so the adapter is unit-testable
+// without constructing a full manager + connection pool.
+type mcpExecutor interface {
+	ExecuteTool(ctx context.Context, prefixedName, inputJSON string) (*tools.MCPToolResult, error)
+}
+
 // mcpAgentTool adapts an tools.MCPTool to the fantasy.AgentTool interface.
 // This keeps the fantasy dependency confined to the agent layer — the tools
 // package is a pure MCP client library with no LLM framework dependency.
 type mcpAgentTool struct {
 	tool            tools.MCPTool
-	manager         *tools.MCPToolManager
+	exec            mcpExecutor
 	providerOptions fantasy.ProviderOptions
 }
 
@@ -29,10 +36,26 @@ func (t *mcpAgentTool) Info() fantasy.ToolInfo {
 }
 
 // Run executes the MCP tool by delegating to the MCPToolManager.
+//
+// MCP-side failures (JSON-RPC protocol errors, transport failures, schema
+// validation rejections from the server) are surfaced to the model as soft
+// tool errors rather than escalated to a critical agent error. This matches
+// the contract that native Kit tools follow via kit.ErrorResult(...) and
+// lets the model self-correct (e.g. retry with a fixed argument shape) or
+// give up gracefully rather than aborting the turn mid-run.
+//
+// Context cancellation is the one exception: if the caller cancelled the
+// context the turn was aborted intentionally, so we propagate the ctx error
+// to let the agent loop unwind cleanly.
 func (t *mcpAgentTool) Run(ctx context.Context, call fantasy.ToolCall) (fantasy.ToolResponse, error) {
-	result, err := t.manager.ExecuteTool(ctx, t.tool.Name, call.Input)
+	result, err := t.exec.ExecuteTool(ctx, t.tool.Name, call.Input)
 	if err != nil {
-		return fantasy.ToolResponse{}, fmt.Errorf("mcp tool execution failed: %w", err)
+		if ctxErr := ctx.Err(); ctxErr != nil {
+			return fantasy.ToolResponse{}, ctxErr
+		}
+		return fantasy.NewTextErrorResponse(
+			fmt.Sprintf("MCP tool %q failed: %s", t.tool.Name, err.Error()),
+		), nil
 	}
 
 	if result.IsError {
@@ -57,8 +80,8 @@ func mcpToolsToAgentTools(mcpTools []tools.MCPTool, manager *tools.MCPToolManage
 	agentTools := make([]fantasy.AgentTool, len(mcpTools))
 	for i, t := range mcpTools {
 		agentTools[i] = &mcpAgentTool{
-			tool:    t,
-			manager: manager,
+			tool: t,
+			exec: manager,
 		}
 	}
 	return agentTools

internal/agent/mcp_adapter_test.go

@@ -0,0 +1,158 @@
+package agent
+
+import (
+	"context"
+	"errors"
+	"strings"
+	"testing"
+	"time"
+
+	"charm.land/fantasy"
+
+	"github.com/mark3labs/kit/internal/tools"
+)
+
+// stubExecutor lets each test script the (result, err) pair returned by
+// ExecuteTool. The adapter holds an mcpExecutor interface, so this is the
+// only seam the tests need.
+type stubExecutor struct {
+	result *tools.MCPToolResult
+	err    error
+	// called records the last invocation for assertion.
+	called bool
+	name   string
+	input  string
+}
+
+func (s *stubExecutor) ExecuteTool(_ context.Context, prefixedName, inputJSON string) (*tools.MCPToolResult, error) {
+	s.called = true
+	s.name = prefixedName
+	s.input = inputJSON
+	return s.result, s.err
+}
+
+func newMCPAgentTool(exec mcpExecutor, name string) *mcpAgentTool {
+	return &mcpAgentTool{
+		tool: tools.MCPTool{Name: name},
+		exec: exec,
+	}
+}
+
+// Manager-side Go errors (JSON-RPC protocol errors, transport failures,
+// schema validation rejections from the MCP server) must be surfaced to
+// the model as soft tool errors so the agent loop can keep going. Aborting
+// the turn would discard all prior tool results — see issue #N.
+func TestMCPAgentTool_RPCErrorBecomesSoftError(t *testing.T) {
+	exec := &stubExecutor{
+		err: errors.New("MCP error -32602: Invalid params: missing field \"task\""),
+	}
+	tool := newMCPAgentTool(exec, "pubmed__search")
+
+	resp, err := tool.Run(context.Background(), fantasy.ToolCall{
+		ID:    "call-1",
+		Name:  "pubmed__search",
+		Input: `{"query":"foo"}`,
+	})
+
+	if err != nil {
+		t.Fatalf("expected nil error (soft), got %v", err)
+	}
+	if !resp.IsError {
+		t.Fatalf("expected IsError=true, got false")
+	}
+	if !strings.Contains(resp.Content, "pubmed__search") {
+		t.Errorf("expected tool name in error content, got %q", resp.Content)
+	}
+	if !strings.Contains(resp.Content, "-32602") {
+		t.Errorf("expected underlying error text in content, got %q", resp.Content)
+	}
+}
+
+// Context cancellation is the one error that must remain critical: it
+// means the caller intentionally aborted, and the agent loop needs to
+// unwind cleanly rather than burning more steps.
+func TestMCPAgentTool_CtxCancelStaysCritical(t *testing.T) {
+	exec := &stubExecutor{
+		// Real managers typically return ctx.Err() (or a wrapper) when the
+		// context is cancelled mid-call.
+		err: context.Canceled,
+	}
+	tool := newMCPAgentTool(exec, "slow__tool")
+
+	ctx, cancel := context.WithCancel(context.Background())
+	cancel()
+
+	resp, err := tool.Run(ctx, fantasy.ToolCall{Name: "slow__tool"})
+
+	if !errors.Is(err, context.Canceled) {
+		t.Fatalf("expected context.Canceled, got %v", err)
+	}
+	if resp.IsError || resp.Content != "" {
+		t.Errorf("expected empty response on critical error, got IsError=%v Content=%q", resp.IsError, resp.Content)
+	}
+}
+
+// Deadline-exceeded behaves the same as cancellation: ctx.Err() is
+// non-nil, so the adapter must propagate the critical error rather than
+// converting the executor's error into a soft response.
+func TestMCPAgentTool_CtxDeadlineStaysCritical(t *testing.T) {
+	exec := &stubExecutor{err: context.DeadlineExceeded}
+	tool := newMCPAgentTool(exec, "slow__tool")
+
+	ctx, cancel := context.WithDeadline(context.Background(), time.Now().Add(-time.Second))
+	defer cancel()
+
+	resp, err := tool.Run(ctx, fantasy.ToolCall{Name: "slow__tool"})
+	if !errors.Is(err, context.DeadlineExceeded) {
+		t.Fatalf("expected context.DeadlineExceeded, got %v", err)
+	}
+	if resp.IsError || resp.Content != "" {
+		t.Errorf("expected empty response on critical error, got IsError=%v Content=%q", resp.IsError, resp.Content)
+	}
+}
+
+// Server-side soft errors (CallToolResult{ isError: true }) must continue
+// to flow through as soft errors — this was the existing behavior and
+// must not regress.
+func TestMCPAgentTool_ServerIsErrorRemainsSoftError(t *testing.T) {
+	exec := &stubExecutor{
+		result: &tools.MCPToolResult{
+			IsError: true,
+			Content: "search service is rate limited; try again in 30s",
+		},
+	}
+	tool := newMCPAgentTool(exec, "pubmed__search")
+
+	resp, err := tool.Run(context.Background(), fantasy.ToolCall{Name: "pubmed__search"})
+	if err != nil {
+		t.Fatalf("expected nil error, got %v", err)
+	}
+	if !resp.IsError {
+		t.Fatalf("expected IsError=true, got false")
+	}
+	if resp.Content != "search service is rate limited; try again in 30s" {
+		t.Errorf("expected pass-through content, got %q", resp.Content)
+	}
+}
+
+// Happy path: ordinary successful tool result is passed through unchanged.
+func TestMCPAgentTool_SuccessIsPassthrough(t *testing.T) {
+	exec := &stubExecutor{
+		result: &tools.MCPToolResult{
+			IsError: false,
+			Content: `{"hits":3}`,
+		},
+	}
+	tool := newMCPAgentTool(exec, "pubmed__search")
+
+	resp, err := tool.Run(context.Background(), fantasy.ToolCall{Name: "pubmed__search"})
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if resp.IsError {
+		t.Fatalf("expected IsError=false")
+	}
+	if resp.Content != `{"hits":3}` {
+		t.Errorf("expected pass-through content, got %q", resp.Content)
+	}
+}

internal/app/app.go

@@ -78,6 +78,13 @@ type App struct {
 	// (~1 frame) so new updates are always let through once the TUI has had a
 	// chance to process the pending event.
 	widgetUpdatePending atomic.Bool
+
+	// steerDrainFn is the test seam used by releaseBusyAfterCompact to pull
+	// any steer messages that arrived during compaction. In production it is
+	// nil and the helper falls back to a.opts.Kit.DrainSteer(); tests that
+	// need to exercise the steer-drain path without standing up a full
+	// *kit.Kit can set this field directly to inject fake items.
+	steerDrainFn func() []queueItem
 }
 
 // New creates a new App with the provided options and pre-loaded messages.
@@ -356,6 +363,10 @@ func (a *App) AddContextMessage(text string) {
 // tea.Program. customInstructions is optional text appended to the summary
 // prompt (e.g. "Focus on the API design decisions").
 //
+// Any prompts queued via Run/RunWithFiles or steering messages injected via
+// Steer/SteerWithFiles while compaction is running are flushed automatically
+// once compaction completes (see releaseBusyAfterCompact).
+//
 // Satisfies ui.AppController.
 func (a *App) CompactConversation(customInstructions string) error {
 	a.mu.Lock()
@@ -377,11 +388,7 @@ func (a *App) CompactConversation(customInstructions string) error {
 
 	go func() {
 		defer a.wg.Done()
-		defer func() {
-			a.mu.Lock()
-			a.busy = false
-			a.mu.Unlock()
-		}()
+		defer a.releaseBusyAfterCompact()
 
 		// Subscribe to SDK events for streaming compaction summary to the TUI.
 		sendFn := func(msg tea.Msg) {
@@ -420,6 +427,9 @@ func (a *App) CompactConversation(customInstructions string) error {
 // CompactAsync is like CompactConversation but calls onComplete/onError
 // callbacks instead of sending TUI events. Used by the extension API's
 // ctx.Compact() which needs callback-based notification.
+//
+// Like CompactConversation, any prompts/steer messages received during
+// compaction are flushed automatically once compaction finishes.
 func (a *App) CompactAsync(customInstructions string, onComplete func(), onError func(string)) error {
 	a.mu.Lock()
 	if a.closed {
@@ -440,11 +450,7 @@ func (a *App) CompactAsync(customInstructions string, onComplete func(), onError
 
 	go func() {
 		defer a.wg.Done()
-		defer func() {
-			a.mu.Lock()
-			a.busy = false
-			a.mu.Unlock()
-		}()
+		defer a.releaseBusyAfterCompact()
 
 		// Subscribe to SDK events for streaming compaction summary to the TUI.
 		sendFn := func(msg tea.Msg) {
@@ -489,6 +495,81 @@ func (a *App) CompactAsync(customInstructions string, onComplete func(), onError
 	return nil
 }
 
+// releaseBusyAfterCompact is the deferred tail that runs at the end of every
+// compaction goroutine (success, error, or panic-after-recover paths). It
+// flips a.busy back to false, but before doing so it checks whether any
+// prompts piled up while compaction was running:
+//
+//   - Run/RunWithFiles append to a.queue when a.busy is set.
+//   - Steer/SteerWithFiles deposit messages into the SDK steer channel via
+//     Kit.InjectSteerWithFiles when a.busy is set.
+//
+// Without this hand-off the queue would sit idle until the user submits
+// another prompt — see issue #27. If we find anything pending we keep busy
+// set, splice the steer messages to the front of the queue, and start a
+// fresh drainQueue goroutine to deliver them as a single batched turn.
+func (a *App) releaseBusyAfterCompact() {
+	// Pull steer messages outside the app mutex; DrainSteer takes its own
+	// internal lock and we don't want to nest the two. The test seam
+	// (a.steerDrainFn) takes precedence so unit tests can inject fake
+	// steer items without a real *kit.Kit.
+	var steerItems []queueItem
+	switch {
+	case a.steerDrainFn != nil:
+		steerItems = a.steerDrainFn()
+	case a.opts.Kit != nil:
+		if leftover := a.opts.Kit.DrainSteer(); len(leftover) > 0 {
+			steerItems = make([]queueItem, len(leftover))
+			for i, sm := range leftover {
+				steerItems[i] = queueItem{Prompt: sm.Text, Files: sm.Files}
+			}
+		}
+	}
+
+	a.mu.Lock()
+	// If the app was closed while compaction was running, drop everything
+	// and just clear busy. Run/Steer would have rejected new items already
+	// after Close(), but this guards against in-flight items that slipped
+	// in just before closed was set.
+	if a.closed {
+		a.queue = a.queue[:0]
+		a.busy = false
+		a.mu.Unlock()
+		return
+	}
+
+	// Combine steer-channel items (front) with the in-memory queue (back).
+	// Steer messages are placed first so they retain their "act now"
+	// semantics relative to ordinary queued prompts that arrived later.
+	pending := append(steerItems, a.queue...)
+	a.queue = a.queue[:0]
+
+	if len(pending) == 0 {
+		a.busy = false
+		a.mu.Unlock()
+		return
+	}
+
+	// Hand off to drainQueue: it will pick up the first item directly and
+	// scoop the rest from a.queue on its first iteration.
+	first := pending[0]
+	if len(pending) > 1 {
+		a.queue = append(a.queue, pending[1:]...)
+	}
+	// Stay busy across the goroutine swap.
+	a.wg.Add(1)
+	a.mu.Unlock()
+
+	// Notify the UI that steer-channel messages were consumed so the
+	// steering badge can clear; ordinary queued prompts will be reflected
+	// by the QueueUpdatedEvent that drainQueue emits as it picks them up.
+	if len(steerItems) > 0 {
+		a.sendEvent(SteerConsumedEvent{})
+	}
+
+	go a.drainQueue(first)
+}
+
 // --------------------------------------------------------------------------
 // Non-interactive execution
 // --------------------------------------------------------------------------

internal/app/app_test.go

@@ -763,3 +763,209 @@ func TestFormatMaxTokensTruncatedMessage_NoKit(t *testing.T) {
 		}
 	}
 }
+
+// --------------------------------------------------------------------------
+// releaseBusyAfterCompact (issue #27)
+// --------------------------------------------------------------------------
+
+// TestReleaseBusyAfterCompact_flushesQueuedMessages is a regression test for
+// issue #27: messages queued via Run() while /compact is running used to sit
+// in a.queue indefinitely until the user typed another prompt. After the fix
+// the deferred releaseBusyAfterCompact tail picks up any pending items and
+// dispatches drainQueue automatically.
+//
+// We simulate the compaction completion path directly (bypassing the SDK)
+// by toggling busy=true, populating the queue exactly as Run() would have
+// during compaction, and then invoking releaseBusyAfterCompact.
+func TestReleaseBusyAfterCompact_flushesQueuedMessages(t *testing.T) {
+	stub := newStubWithFuncs(
+		func(ctx context.Context) (*kit.TurnResult, error) {
+			return turnResult("compacted then drained"), nil
+		},
+	)
+	app := newTestApp(stub)
+	defer app.Close()
+
+	// Simulate the state at the start of the compaction tail: busy is set
+	// and a couple of prompts have piled up in the queue while we were
+	// summarising. (Run() would have appended them and returned a queue
+	// length > 0 to the caller.)
+	app.mu.Lock()
+	app.busy = true
+	app.queue = append(app.queue,
+		queueItem{Prompt: "queued during compact #1"},
+		queueItem{Prompt: "queued during compact #2"},
+	)
+	app.mu.Unlock()
+
+	// Invoke the deferred tail directly. It should kick off drainQueue.
+	app.releaseBusyAfterCompact()
+
+	// drainQueue runs in a goroutine. Wait for the app to come back to idle.
+	ok := waitForCondition(2*time.Second, func() bool {
+		app.mu.Lock()
+		defer app.mu.Unlock()
+		return !app.busy
+	})
+	if !ok {
+		t.Fatal("app did not become idle after releaseBusyAfterCompact: queue not drained")
+	}
+
+	// Wait for any in-flight goroutine to finish before reading state.
+	app.wg.Wait()
+
+	if got := app.QueueLength(); got != 0 {
+		t.Fatalf("expected empty queue after drain, got %d", got)
+	}
+	if n := stub.callCount(); n == 0 {
+		t.Fatalf("expected stub PromptFunc to fire at least once after compact, got %d calls", n)
+	}
+}
+
+// TestReleaseBusyAfterCompact_idleWhenQueueEmpty verifies that with no
+// pending messages the helper just clears busy and does NOT spawn a
+// drainQueue goroutine (no spurious agent turn).
+func TestReleaseBusyAfterCompact_idleWhenQueueEmpty(t *testing.T) {
+	stub := newStub()
+	app := newTestApp(stub)
+	defer app.Close()
+
+	app.mu.Lock()
+	app.busy = true
+	app.mu.Unlock()
+
+	app.releaseBusyAfterCompact()
+
+	app.mu.Lock()
+	busy := app.busy
+	app.mu.Unlock()
+	if busy {
+		t.Fatal("expected busy=false after releaseBusyAfterCompact with empty queue")
+	}
+
+	// Give any rogue goroutine a moment to (incorrectly) call PromptFunc.
+	time.Sleep(50 * time.Millisecond)
+	if n := stub.callCount(); n != 0 {
+		t.Fatalf("expected 0 PromptFunc calls when queue empty, got %d", n)
+	}
+}
+
+// TestReleaseBusyAfterCompact_splicesSteerAheadOfQueue exercises the SDK
+// steer-drain branch of releaseBusyAfterCompact (issue #27 follow-up).
+//
+// Production wires a.opts.Kit.DrainSteer() to pull messages that arrived via
+// Steer/SteerWithFiles during compaction, but Options.Kit is *kit.Kit (a
+// concrete struct) so unit tests cannot stand up a real instance without a
+// full LLM backend. The test uses the unexported steerDrainFn seam to inject
+// fake steer items, then asserts that:
+//
+//   - Steer items are dispatched ahead of any prompts that piled up in
+//     a.queue (steer retains "act now" priority over ordinary queued
+//     prompts), and
+//   - the helper still hands off to drainQueue so the steer item actually
+//     fires (the previous behaviour left them stranded — see #27).
+func TestReleaseBusyAfterCompact_splicesSteerAheadOfQueue(t *testing.T) {
+	var pmu sync.Mutex
+	var firstPrompt string
+	stub := newStubWithFuncs(
+		func(ctx context.Context) (*kit.TurnResult, error) {
+			return turnResult("steer dispatched"), nil
+		},
+	)
+	// Wrap PromptFunc so we can capture the prompt text the stub receives
+	// (newStubWithFuncs's fns ignore prompt; we need it to verify ordering).
+	capturingPrompt := func(ctx context.Context, prompt string) (*kit.TurnResult, error) {
+		pmu.Lock()
+		if firstPrompt == "" {
+			firstPrompt = prompt
+		}
+		pmu.Unlock()
+		return stub.fn(ctx, prompt)
+	}
+	app := New(Options{PromptFunc: capturingPrompt}, nil)
+	defer app.Close()
+
+	// Inject fake steer items via the test seam. In production the same
+	// items would have been delivered through Kit.InjectSteerWithFiles
+	// during /compact and pulled by DrainSteer here.
+	app.steerDrainFn = func() []queueItem {
+		return []queueItem{
+			{Prompt: "steer-1"},
+			{Prompt: "steer-2"},
+		}
+	}
+
+	// Simulate the state at the end of compaction: busy is set and a couple
+	// of regular Run() prompts have piled up after the steer messages.
+	app.mu.Lock()
+	app.busy = true
+	app.queue = append(app.queue,
+		queueItem{Prompt: "queued-1"},
+		queueItem{Prompt: "queued-2"},
+	)
+	app.mu.Unlock()
+
+	app.releaseBusyAfterCompact()
+
+	// Wait for the dispatched batch to complete.
+	ok := waitForCondition(2*time.Second, func() bool {
+		app.mu.Lock()
+		defer app.mu.Unlock()
+		return !app.busy
+	})
+	if !ok {
+		t.Fatal("app did not become idle after steer-spliced releaseBusyAfterCompact")
+	}
+	app.wg.Wait()
+
+	// drainQueue picks up `first` directly and batches the rest. With
+	// PromptFunc set, executeBatch invokes us with items[0] only — that
+	// item must be the first steer message, proving steer items were
+	// spliced ahead of the previously queued prompts.
+	pmu.Lock()
+	got := firstPrompt
+	pmu.Unlock()
+	if got != "steer-1" {
+		t.Fatalf("expected first dispatched prompt to be steer item %q (steer items must come before queued prompts), got %q",
+			"steer-1", got)
+	}
+
+	// Queue should be fully drained and PromptFunc must have actually fired.
+	if n := app.QueueLength(); n != 0 {
+		t.Fatalf("expected empty queue after drain, got %d entries", n)
+	}
+	if n := stub.callCount(); n == 0 {
+		t.Fatal("expected stub PromptFunc to fire at least once after splice")
+	}
+}
+
+// TestReleaseBusyAfterCompact_dropsQueueWhenClosed verifies that if the app
+// was closed during compaction the helper discards any pending items rather
+// than spawning drainQueue against a torn-down App.
+func TestReleaseBusyAfterCompact_dropsQueueWhenClosed(t *testing.T) {
+	stub := newStub()
+	app := newTestApp(stub)
+
+	app.mu.Lock()
+	app.busy = true
+	app.queue = append(app.queue, queueItem{Prompt: "would have run"})
+	app.closed = true
+	app.mu.Unlock()
+
+	app.releaseBusyAfterCompact()
+
+	app.mu.Lock()
+	busy := app.busy
+	qLen := len(app.queue)
+	app.mu.Unlock()
+	if busy {
+		t.Fatal("expected busy=false even when closed")
+	}
+	if qLen != 0 {
+		t.Fatalf("expected queue cleared on closed app, got %d entries", qLen)
+	}
+	time.Sleep(20 * time.Millisecond)
+	if n := stub.callCount(); n != 0 {
+		t.Fatalf("expected 0 PromptFunc calls on closed app, got %d", n)
+	}
+}

internal/session/compaction_test.go

@@ -129,26 +129,35 @@ func TestCompactionWithNewMessagesAfterCompaction(t *testing.T) {
 	msg4 := message.Message{Role: message.RoleAssistant, Parts: []message.ContentPart{message.TextContent{Text: "Message 4 - after compaction"}}}
 	_, _ = tm.AppendMessage(msg4)
 
-	// BuildContext should return: [summary] + [M4 (new after compaction)] + [M3 (kept)]
+	// BuildContext should return: [summary] + [M3 (kept)] + [M4 (new after compaction)]
+	// Kept messages must appear BEFORE post-compaction messages so the LLM
+	// sees the conversation in chronological order. Otherwise the latest
+	// post-compaction user message would be followed by an older kept user
+	// message, breaking user/assistant alternation and causing the model to
+	// respond as if the post-compaction turn never happened.
 	messages, _, _ := tm.BuildContext()
 	if len(messages) != 3 {
-		t.Fatalf("expected 3 messages (summary + M4 + M3), got %d: %+v", len(messages), messages)
+		t.Fatalf("expected 3 messages (summary + M3 + M4), got %d: %+v", len(messages), messages)
 	}
 
-	// Verify order: summary, M4 (new), M3 (kept)
+	// Verify order: summary, M3 (kept), M4 (new)
 	if messages[0].Role != fantasy.MessageRoleSystem {
 		t.Errorf("first message should be summary, got %s", messages[0].Role)
 	}
-	if messages[1].Role != fantasy.MessageRoleAssistant {
-		t.Errorf("second message should be assistant (M4), got %s", messages[1].Role)
+	if messages[1].Role != fantasy.MessageRoleUser {
+		t.Errorf("second message should be user (M3 kept), got %s", messages[1].Role)
 	}
-	m4Text := messages[1].Content[0].(fantasy.TextPart).Text
+	m3Text := messages[1].Content[0].(fantasy.TextPart).Text
+	if m3Text != "Message 3 - kept" {
+		t.Errorf("unexpected M3 text: %s", m3Text)
+	}
+	if messages[2].Role != fantasy.MessageRoleAssistant {
+		t.Errorf("third message should be assistant (M4 post-compact), got %s", messages[2].Role)
+	}
+	m4Text := messages[2].Content[0].(fantasy.TextPart).Text
 	if m4Text != "Message 4 - after compaction" {
 		t.Errorf("unexpected M4 text: %s", m4Text)
 	}
-	if messages[2].Role != fantasy.MessageRoleUser {
-		t.Errorf("third message should be user (M3), got %s", messages[2].Role)
-	}
 
 	// Verify that M1 is NOT in the context
 	for i, msg := range messages {

internal/session/tree_manager.go

@@ -755,9 +755,17 @@ func (tm *TreeManager) BuildContext() (messages []fantasy.Message, provider stri
 		}
 	}
 
-	// If there is a compaction, inject the summary first and collect
-	// the kept messages starting from FirstKeptEntryID (since the
-	// compaction entry's parent chain doesn't include them).
+	// If there is a compaction, inject the summary first, then the
+	// preserved "kept" messages (chronologically before the compaction),
+	// then the post-compaction messages (chronologically after).
+	//
+	// Order matters: the kept messages must come BEFORE the post-compaction
+	// branch so the LLM sees the conversation in chronological order. If the
+	// kept messages were appended last, the latest user message in the
+	// current branch would be followed by an older kept user message,
+	// breaking the strict user/assistant alternation that providers expect
+	// and causing the model to respond as if the previous turn never
+	// happened.
 	if lastCompaction != nil {
 		messages = append(messages, fantasy.Message{
 			Role: fantasy.MessageRoleSystem,
@@ -768,49 +776,10 @@ func (tm *TreeManager) BuildContext() (messages []fantasy.Message, provider stri
 			},
 		})
 
-		// Collect entries from the compaction entry itself (at compactionIndex)
-		// and any entries before it in the branch (newer messages).
-		for i := compactionIndex; i < len(branch); i++ {
-			entry := branch[i]
-			switch e := entry.(type) {
-			case *MessageEntry:
-				msg, err := e.ToMessage()
-				if err != nil {
-					continue // skip malformed entries
-				}
-				msgs := msg.ToLLMMessages()
-				messages = append(messages, msgs...)
-
-			case *BranchSummaryEntry:
-				// Convert branch summary to a user message for context.
-				if e.Summary != "" {
-					messages = append(messages, fantasy.Message{
-						Role: fantasy.MessageRoleUser,
-						Content: []fantasy.MessagePart{
-							fantasy.TextPart{
-								Text: fmt.Sprintf("[Branch context: %s]", e.Summary),
-							},
-						},
-					})
-				}
-
-			case *ModelChangeEntry:
-				provider = e.Provider
-				modelID = e.ModelID
-
-			case *CompactionEntry:
-				// Already handled above (summary injected).
-				continue
-			}
-		}
-
-		// Now collect the kept messages starting from FirstKeptEntryID.
-		// These are not in the current branch because the compaction entry
-		// is parented to the first kept entry's parent, not the first kept entry.
-		// We iterate through entries in order (not using getBranchLocked) to avoid
-		// walking back to old compacted messages.
-		// We stop when we reach the compaction entry to avoid double-counting
-		// messages that were added after the compaction.
+		// Step 1: collect the kept messages starting from FirstKeptEntryID.
+		// These are not on the current branch (the compaction entry is a
+		// new root with no parent), so we iterate tm.entries in append order
+		// and stop when we reach the compaction entry itself.
 		if lastCompaction.FirstKeptEntryID != "" {
 			found := false
 			for _, entry := range tm.entries {
@@ -825,13 +794,12 @@ func (tm *TreeManager) BuildContext() (messages []fantasy.Message, provider stri
 					}
 				}
 
-				// Stop when we reach the compaction entry itself.
-				// Messages after the compaction are collected from the branch walk above.
+				// Stop when we reach the compaction entry itself; messages
+				// after it are collected from the branch walk below.
 				if entryID == lastCompaction.ID {
 					break
 				}
 
-				// Process this kept entry.
 				switch e := entry.(type) {
 				case *MessageEntry:
 					msg, err := e.ToMessage()
@@ -860,6 +828,42 @@ func (tm *TreeManager) BuildContext() (messages []fantasy.Message, provider stri
 			}
 		}
 
+		// Step 2: collect entries on the current branch after the compaction
+		// entry (these are post-compaction messages). The compaction entry
+		// itself is skipped — its summary was already injected above.
+		for i := compactionIndex; i < len(branch); i++ {
+			entry := branch[i]
+			switch e := entry.(type) {
+			case *MessageEntry:
+				msg, err := e.ToMessage()
+				if err != nil {
+					continue
+				}
+				msgs := msg.ToLLMMessages()
+				messages = append(messages, msgs...)
+
+			case *BranchSummaryEntry:
+				if e.Summary != "" {
+					messages = append(messages, fantasy.Message{
+						Role: fantasy.MessageRoleUser,
+						Content: []fantasy.MessagePart{
+							fantasy.TextPart{
+								Text: fmt.Sprintf("[Branch context: %s]", e.Summary),
+							},
+						},
+					})
+				}
+
+			case *ModelChangeEntry:
+				provider = e.Provider
+				modelID = e.ModelID
+
+			case *CompactionEntry:
+				// Summary already injected above.
+				continue
+			}
+		}
+
 		return messages, provider, modelID
 	}
 
@@ -1030,44 +1034,22 @@ func (tm *TreeManager) GetContextEntryIDs() []string {
 
 	var ids []string
 
-	// If there's a compaction, we need to collect IDs from:
-	// 1. Entries after the compaction entry in the branch (newer messages)
-	// 2. Entries from FirstKeptEntryID onwards (kept messages)
+	// If there's a compaction, we collect IDs in the same order as
+	// BuildContext: [summary placeholder, kept messages, post-compaction
+	// messages]. This ordering must stay in sync with BuildContext so a
+	// cut-point index can be mapped back to the correct entry ID.
 	if lastCompaction != nil {
 		// Placeholder for the summary system message (no entry ID).
 		ids = append(ids, "")
 
-		// Collect IDs from entries after the compaction entry (newer messages).
-		for i := compactionIndex + 1; i < len(branch); i++ {
-			entry := branch[i]
-			switch e := entry.(type) {
-			case *MessageEntry:
-				msg, err := e.ToMessage()
-				if err != nil {
-					continue
-				}
-				msgs := msg.ToLLMMessages()
-				for range msgs {
-					ids = append(ids, e.ID)
-				}
-
-			case *BranchSummaryEntry:
-				if e.Summary != "" {
-					ids = append(ids, e.ID)
-				}
-			}
-		}
-
-		// Collect IDs from the kept messages starting at FirstKeptEntryID.
-		// We iterate through entries in order (not using getBranchLocked) to avoid
-		// walking back to old compacted messages.
-		// We stop when we reach the compaction entry to avoid double-counting.
+		// Step 1: IDs of the kept messages starting at FirstKeptEntryID.
+		// Iterate tm.entries in append order and stop at the compaction
+		// entry to avoid double-counting post-compaction messages.
 		if lastCompaction.FirstKeptEntryID != "" {
 			found := false
 			for _, entry := range tm.entries {
 				entryID := tm.EntryID(entry)
 
-				// Skip entries until we reach the first kept entry.
 				if !found {
 					if entryID == lastCompaction.FirstKeptEntryID {
 						found = true
@@ -1076,7 +1058,6 @@ func (tm *TreeManager) GetContextEntryIDs() []string {
 					}
 				}
 
-				// Stop when we reach the compaction entry itself.
 				if entryID == lastCompaction.ID {
 					break
 				}
@@ -1100,6 +1081,28 @@ func (tm *TreeManager) GetContextEntryIDs() []string {
 			}
 		}
 
+		// Step 2: IDs of entries after the compaction entry on the current
+		// branch (post-compaction messages).
+		for i := compactionIndex + 1; i < len(branch); i++ {
+			entry := branch[i]
+			switch e := entry.(type) {
+			case *MessageEntry:
+				msg, err := e.ToMessage()
+				if err != nil {
+					continue
+				}
+				msgs := msg.ToLLMMessages()
+				for range msgs {
+					ids = append(ids, e.ID)
+				}
+
+			case *BranchSummaryEntry:
+				if e.Summary != "" {
+					ids = append(ids, e.ID)
+				}
+			}
+		}
+
 		return ids
 	}
 

internal/ui/commands/commands.go

@@ -161,6 +161,12 @@ var SlashCommands = []SlashCommand{
 		Category:    "Navigation",
 		Aliases:     []string{"/r"},
 	},
+	{
+		Name:        "/copy",
+		Description: "Copy the last message to the system clipboard",
+		Category:    "System",
+		Aliases:     []string{"/cp"},
+	},
 	{
 		Name:        "/export",
 		Description: "Export session (JSONL by default, or /export path.jsonl)",

internal/ui/model.go

@@ -129,8 +129,18 @@ type AppController interface {
 // SkillItem holds display metadata about a loaded skill for the startup
 // [Skills] section. Built by the CLI layer from the SDK's []*kit.Skill.
 type SkillItem struct {
-	Name   string // Skill name (e.g. "btca-cli").
-	Path   string // Absolute path to the skill file.
+	Name        string // Skill name (e.g. "btca-cli").
+	Path        string // Absolute path to the skill file.
+	Source      string // "project" or "user" (global).
+	Description string // Short summary used in autocomplete and help.
+}
+
+// ExtensionItem holds display metadata about a loaded extension for the
+// startup [Extensions] section. Built by the CLI layer from the SDK's
+// []kit.ExtensionInfo.
+type ExtensionItem struct {
+	Name   string // Extension display name (filename without .go extension).
+	Path   string // Absolute path to the extension's .go file.
 	Source string // "project" or "user" (global).
 }
 
@@ -363,6 +373,16 @@ type AppModelOptions struct {
 	// watcher detects changes. May be nil if skill hot-reload is not needed.
 	GetSkillItems func() []SkillItem
 
+	// ExtensionItems lists loaded extensions for the [Extensions] startup
+	// section. Each entry shows the filename of an extension that was
+	// discovered and loaded (global, project-local, or explicit).
+	ExtensionItems []ExtensionItem
+
+	// GetExtensionItems, if non-nil, returns the current extension items.
+	// Called on extension hot-reload to refresh the list. May be nil if no
+	// extensions are loaded.
+	GetExtensionItems func() []ExtensionItem
+
 	// MCPToolCount is the number of tools loaded from external MCP servers.
 	MCPToolCount int
 
@@ -607,6 +627,14 @@ type AppModel struct {
 	// skill list after content hot-reload. May be nil.
 	getSkillItems func() []SkillItem
 
+	// extensionItems lists loaded extensions for the [Extensions] startup
+	// section (filenames only).
+	extensionItems []ExtensionItem
+
+	// getExtensionItems returns the current extension items. Used to refresh
+	// the list after extension hot-reload. May be nil.
+	getExtensionItems func() []ExtensionItem
+
 	// mcpToolCount and extensionToolCount track tool counts by source for
 	// the startup info display.
 	mcpToolCount       int
@@ -860,6 +888,8 @@ func NewAppModel(appCtrl AppController, opts AppModelOptions) *AppModel {
 	m.contextPaths = opts.ContextPaths
 	m.skillItems = opts.SkillItems
 	m.getSkillItems = opts.GetSkillItems
+	m.extensionItems = opts.ExtensionItems
+	m.getExtensionItems = opts.GetExtensionItems
 	m.mcpToolCount = opts.MCPToolCount
 	m.extensionToolCount = opts.ExtensionToolCount
 	m.startupExtensionMessages = opts.StartupExtensionMessages
@@ -912,6 +942,20 @@ func NewAppModel(appCtrl AppController, opts AppModelOptions) *AppModel {
 		}
 	}
 
+	// Merge skills into autocomplete as /skill:<name> commands. Skills accept
+	// optional trailing args, so HasArgs is true — Enter populates the input
+	// with "/skill:name " rather than auto-submitting.
+	if ic, ok := m.input.(*InputComponent); ok && len(opts.SkillItems) > 0 {
+		for _, s := range opts.SkillItems {
+			ic.commands = append(ic.commands, commands.SlashCommand{
+				Name:        "/skill:" + s.Name,
+				Description: formatSkillDescription(s),
+				Category:    "Skills",
+				HasArgs:     true,
+			})
+		}
+	}
+
 	// Merge MCP prompts into autocomplete as /<server>:<prompt> commands.
 	if ic, ok := m.input.(*InputComponent); ok && len(opts.MCPPrompts) > 0 {
 		for _, p := range opts.MCPPrompts {
@@ -1014,8 +1058,21 @@ func (m *AppModel) AddStartupMessageToScrollList() {
 		pairs = append(pairs, [2]string{"Skills", strings.Join(names, ", ")})
 	}
 
-	// Extension tool count (only shown when > 0).
-	if m.extensionToolCount > 0 {
+	// Extensions — listed by filename. Each extension shows its basename
+	// without the .go suffix, matching the [Skills] section's style.
+	if len(m.extensionItems) > 0 {
+		names := make([]string, len(m.extensionItems))
+		for i, ei := range m.extensionItems {
+			names[i] = ei.Name
+		}
+		value := strings.Join(names, ", ")
+		if m.extensionToolCount > 0 {
+			value += fmt.Sprintf(" (%d tools)", m.extensionToolCount)
+		}
+		pairs = append(pairs, [2]string{"Extensions", value})
+	} else if m.extensionToolCount > 0 {
+		// Fallback: tool count only (extensions registered tools but the CLI
+		// did not provide ExtensionItems for some reason).
 		pairs = append(pairs, [2]string{"Extensions", fmt.Sprintf("%d tools", m.extensionToolCount)})
 	}
 
@@ -1251,7 +1308,11 @@ func (m *AppModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 			m.scrollList.autoScroll = false
 		case tea.MouseWheelDown:
 			m.scrollList.ScrollBy(scrollLines)
-			if m.scrollList.AtBottom() {
+			// Only re-enable auto-scroll when the user is not actively
+			// selecting text. Otherwise a wheel-down during a drag-select
+			// would re-arm GotoBottom on the next stream chunk, shifting
+			// the highlighted row out from under the cursor.
+			if m.scrollList.AtBottom() && !m.scrollList.IsMouseDown() {
 				m.scrollList.autoScroll = true
 			}
 		}
@@ -1259,9 +1320,14 @@ func (m *AppModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 	// ── Mouse click selection (crush-style character-level) ──────────────────
 	case tea.MouseClickMsg:
 		if msg.Button == tea.MouseLeft {
-			// Calculate viewport-relative coordinates.
-			viewY := msg.Y - m.scrollbackYOffset
-			if viewY >= 0 && viewY < m.scrollList.height {
+			// Compute the scrollback origin from the current frame's layout
+			// rather than the stale cached value from the previous View().
+			// scrollbackYOffset/scrollList.height are only refreshed inside
+			// View() and lag behind any state change that resized the header
+			// (extension widgets, warning rows, etc.) since the last render.
+			yOff, vpHeight := m.currentScrollbackBounds()
+			viewY := msg.Y - yOff
+			if viewY >= 0 && viewY < vpHeight {
 				// Clear any previous selection on a new click.
 				// HandleMouseDown will set up new selection state.
 				if m.scrollList.HandleMouseDown(msg.X, viewY) {
@@ -1272,8 +1338,9 @@ func (m *AppModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 
 	// ── Mouse motion/drag for character-level selection ──────────────────────
 	case tea.MouseMotionMsg:
-		viewY := msg.Y - m.scrollbackYOffset
-		if viewY >= 0 && viewY < m.scrollList.height {
+		yOff, vpHeight := m.currentScrollbackBounds()
+		viewY := msg.Y - yOff
+		if viewY >= 0 && viewY < vpHeight {
 			m.scrollList.HandleMouseDrag(msg.X, viewY)
 		}
 
@@ -1603,10 +1670,16 @@ func (m *AppModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 
 	// ── Cancel timer expired ─────────────────────────────────────────────────
 	case uicore.CancelTimerExpiredMsg:
+		if m.canceling {
+			m.layoutDirty = true
+		}
 		m.canceling = false
 
 	// ── Ctrl+C reset timer expired ────────────────────────────────────────────
 	case uicore.CtrlCResetMsg:
+		if m.ctrlCPressedOnce {
+			m.layoutDirty = true
+		}
 		m.ctrlCPressedOnce = false
 
 	// ── Input submitted ──────────────────────────────────────────────────────
@@ -2328,6 +2401,7 @@ func (m *AppModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 		if msg.err != nil {
 			m.printSystemMessage(fmt.Sprintf("Extension reload failed: %v", msg.err))
 		} else {
+			m.refreshExtensionItems()
 			m.printSystemMessage("Extensions reloaded.")
 		}
 
@@ -3095,6 +3169,8 @@ func (m *AppModel) handleSlashCommand(sc *commands.SlashCommand, args string) te
 		return m.handleResumeCommand()
 	case "/export":
 		return m.handleExportCommand(args)
+	case "/copy":
+		return m.handleCopyCommand()
 	case "/share":
 		return m.handleShareCommand()
 	case "/import":
@@ -3395,13 +3471,56 @@ func (m *AppModel) refreshPromptTemplates() {
 	}
 }
 
-// refreshSkillItems reloads skill items from the provider callback.
-// Called on ContentReloadEvent.
+// refreshSkillItems reloads skill items from the provider callback and
+// updates the autocomplete entries. Called on ContentReloadEvent.
 func (m *AppModel) refreshSkillItems() {
 	if m.getSkillItems == nil {
 		return
 	}
-	m.skillItems = m.getSkillItems()
+	newItems := m.getSkillItems()
+	m.skillItems = newItems
+
+	if ic, ok := m.input.(*InputComponent); ok {
+		// Remove old Skills commands and add fresh ones.
+		var kept []commands.SlashCommand
+		for _, sc := range ic.commands {
+			if sc.Category != "Skills" {
+				kept = append(kept, sc)
+			}
+		}
+		for _, s := range newItems {
+			kept = append(kept, commands.SlashCommand{
+				Name:        "/skill:" + s.Name,
+				Description: formatSkillDescription(s),
+				Category:    "Skills",
+				HasArgs:     true,
+			})
+		}
+		ic.commands = kept
+	}
+}
+
+// refreshExtensionItems reloads extension items from the provider callback
+// so the [Extensions] startup section reflects the current set after a
+// hot-reload. Called from the extReloadResultMsg handler.
+func (m *AppModel) refreshExtensionItems() {
+	if m.getExtensionItems == nil {
+		return
+	}
+	m.extensionItems = m.getExtensionItems()
+}
+
+// formatSkillDescription returns the autocomplete description for a skill,
+// prefixed with [project] or [user] so users can tell colliding names apart.
+func formatSkillDescription(s SkillItem) string {
+	prefix := "[user]"
+	if s.Source == "project" {
+		prefix = "[project]"
+	}
+	if s.Description == "" {
+		return prefix
+	}
+	return prefix + " " + s.Description
 }
 
 // refreshMCPPrompts reloads MCP prompts from the provider callback and
@@ -3476,6 +3595,7 @@ func (m *AppModel) printHelpMessage() {
 		"**System:**\n" +
 		"- `/compact [instructions]`: Summarise older messages to free context space\n" +
 		"- `/clear`: Clear message history\n" +
+		"- `/copy`: Copy the last message to the system clipboard\n" +
 		"- `/export [path]`: Export session as JSONL\n" +
 		"- `/import <path.jsonl>`: Import session from JSONL file\n" +
 		"- `/reset-usage`: Reset usage statistics\n" +
@@ -3712,7 +3832,12 @@ func (m *AppModel) appendStreamingChunk(role, content string) {
 		}
 		// Auto-scroll to bottom if enabled (iteratr pattern)
 		// Don't call SetItems() - the slice reference hasn't changed
-		if m.scrollList != nil {
+		//
+		// CRITICAL: never scroll the viewport while the user is actively
+		// selecting text (mouse button held). Doing so shifts the
+		// highlighted content out from under the cursor and produces the
+		// off-by-N-row drift users see when copy-selecting during streaming.
+		if m.scrollList != nil && !m.scrollList.IsMouseDown() {
 			if m.scrollList.autoScroll {
 				m.scrollList.GotoBottom()
 			} else if m.scrollList.AtBottom() {
@@ -3740,6 +3865,36 @@ func (m *AppModel) appendStreamingChunk(role, content string) {
 	m.refreshContent()
 }
 
+// currentScrollbackBounds returns the live (yOffset, viewportHeight) for the
+// scrollback region, computed from the current state — not from the cached
+// values populated inside View().
+//
+// scrollbackYOffset and scrollList.height are refreshed once per render, so
+// any state change that resizes the header (extension widget toggles,
+// warning rows, queued messages, etc.) leaves the cached values one frame
+// stale. Mouse click handlers in Update() can then place the cursor on the
+// wrong line, producing the off-by-N-row drift seen during copy-selection.
+//
+// This recomputes the header height by rendering it (cheap — the renderer
+// returns "" when no extension header is set) and recomputes the viewport
+// height the same way distributeHeight() does, so both inputs to the
+// y → (item, line) mapping are always current.
+func (m *AppModel) currentScrollbackBounds() (yOffset, viewportHeight int) {
+	// Force a fresh layout if anything in Update() marked the state dirty;
+	// otherwise scrollList.height still reflects the previous frame.
+	if m.layoutDirty {
+		m.distributeHeight()
+		m.layoutDirty = false
+	}
+	if headerView := m.renderHeaderFooter(m.getHeader); headerView != "" {
+		yOffset = lipgloss.Height(headerView)
+	}
+	if m.scrollList != nil {
+		viewportHeight = m.scrollList.height
+	}
+	return yOffset, viewportHeight
+}
+
 // distributeHeight recalculates child component heights after a window resize,
 // queue change, widget update, or state transition, and propagates the computed
 // stream height to the StreamComponent.
@@ -3812,7 +3967,20 @@ func (m *AppModel) distributeHeight() {
 		headerFooterLines += lipgloss.Height(footerView)
 	}
 
-	streamHeight := max(m.height-separatorLines-widgetLines-headerFooterLines-queuedLines-inputLines-statusBarLines, 0)
+	// Account for transient warning rows that View() injects between the
+	// scrollback and the separator. These flags are toggled by ESC/Ctrl+C
+	// handlers; without subtracting them here the joined view exceeds
+	// m.height by one line per active warning and the bottom of the screen
+	// gets silently clipped — which in turn invalidates scrollbackYOffset.
+	var warningLines int
+	if m.canceling {
+		warningLines++
+	}
+	if m.ctrlCPressedOnce {
+		warningLines++
+	}
+
+	streamHeight := max(m.height-separatorLines-widgetLines-headerFooterLines-queuedLines-inputLines-statusBarLines-warningLines, 0)
 
 	// In alt screen mode, give the calculated height to ScrollList instead of stream.
 	// The stream component still exists but is embedded as the last item in scrollList.
@@ -4236,6 +4404,48 @@ func (m *AppModel) handleNameCommand(args string) tea.Cmd {
 	return nil
 }
 
+// handleCopyCommand copies the last user or assistant message to the system
+// clipboard. Skips transient system messages (e.g. /help output) so the user
+// gets the actual last conversational message.
+func (m *AppModel) handleCopyCommand() tea.Cmd {
+	if len(m.messages) == 0 {
+		m.printSystemMessage("No messages to copy.")
+		return nil
+	}
+
+	var (
+		text string
+		role string
+	)
+	for i := len(m.messages) - 1; i >= 0; i-- {
+		switch msg := m.messages[i].(type) {
+		case *TextMessageItem:
+			if msg.role == "user" || msg.role == "assistant" {
+				text = msg.content
+				role = msg.role
+			}
+		case *StreamingMessageItem:
+			if msg.role == "assistant" || msg.role == "reasoning" {
+				text = msg.content.String()
+				role = msg.role
+			}
+		}
+		if text != "" {
+			break
+		}
+	}
+
+	if strings.TrimSpace(text) == "" {
+		m.printSystemMessage("No copyable message found.")
+		return nil
+	}
+
+	m.printSystemMessage(fmt.Sprintf(
+		"Copied last %s message to clipboard (%d chars).", role, len(text),
+	))
+	return clipboard.CopyToClipboard(text)
+}
+
 // handleExportCommand exports the current session to a file.
 // Usage: /export          — copies the JSONL file to cwd with a descriptive name.
 //

internal/ui/scrolllist.go

@@ -60,10 +60,13 @@ func NewScrollList(width, height int) *ScrollList {
 }
 
 // SetItems replaces the items in the scroll list. If auto-scroll is enabled,
-// the viewport will scroll to the bottom to show the latest content.
+// the viewport will scroll to the bottom to show the latest content — EXCEPT
+// when the user is actively selecting text (mouse button held), in which case
+// the scroll position is locked so the highlighted content stays under the
+// cursor. The pending bottom-scroll is deferred to MouseUp.
 func (s *ScrollList) SetItems(items []MessageItem) {
 	s.items = items
-	if s.autoScroll {
+	if s.autoScroll && !s.sel.MouseDown {
 		s.GotoBottom()
 	}
 }
@@ -157,6 +160,10 @@ func (s *ScrollList) HandleMouseDown(x, y int) bool {
 // HandleMouseDrag handles mouse motion while button is held.
 // Updates the selection endpoint for character-level precision.
 // Returns true if selection was updated.
+//
+// Defensively disables auto-scroll on every drag update — even if the
+// MouseDown handler missed (e.g. click landed in viewport padding), any
+// active drag means the user is selecting and the viewport must not jump.
 func (s *ScrollList) HandleMouseDrag(x, y int) bool {
 	if !s.sel.MouseDown {
 		return false
@@ -171,6 +178,9 @@ func (s *ScrollList) HandleMouseDrag(x, y int) bool {
 		return false
 	}
 
+	// Hard-lock the viewport while dragging.
+	s.autoScroll = false
+
 	s.sel.DragItemIdx = itemIdx
 	s.sel.DragLineIdx = lineIdx
 	s.sel.DragCol = x
@@ -178,6 +188,13 @@ func (s *ScrollList) HandleMouseDrag(x, y int) bool {
 	return true
 }
 
+// IsMouseDown reports whether the user currently has the mouse button held
+// (i.e. a selection drag is in progress). Used by the parent model to avoid
+// re-enabling auto-scroll during streaming while the user is selecting.
+func (s *ScrollList) IsMouseDown() bool {
+	return s.sel.MouseDown
+}
+
 // HandleMouseUp handles mouse button release.
 // Returns true if there was an active selection.
 func (s *ScrollList) HandleMouseUp() bool {

internal/ui/scrolllist_test.go

@@ -0,0 +1,132 @@
+package ui
+
+import (
+	"fmt"
+	"strings"
+	"testing"
+)
+
+// fakeItem is a deterministic MessageItem for ScrollList tests.
+type fakeItem struct {
+	id    string
+	lines int
+}
+
+func (f *fakeItem) ID() string { return f.id }
+func (f *fakeItem) Render(_ int) string {
+	if f.lines <= 0 {
+		return ""
+	}
+	parts := make([]string, f.lines)
+	for i := range parts {
+		parts[i] = fmt.Sprintf("%s-line-%d", f.id, i)
+	}
+	return strings.Join(parts, "\n")
+}
+func (f *fakeItem) Height() int { return f.lines }
+
+// makeItems builds n fake items of `lines` height each.
+func makeItems(n, lines int) []MessageItem {
+	out := make([]MessageItem, n)
+	for i := range out {
+		out[i] = &fakeItem{id: fmt.Sprintf("item-%d", i), lines: lines}
+	}
+	return out
+}
+
+// TestScrollList_MouseDownPreventsAutoScroll verifies the core fix for the
+// copy-selection drift bug: while the user has the mouse button held
+// (drag-selecting), incoming content updates must NOT shift the viewport,
+// because doing so moves the highlighted content out from under the cursor.
+func TestScrollList_MouseDownPreventsAutoScroll(t *testing.T) {
+	sl := NewScrollList(80, 10)
+	sl.SetItems(makeItems(20, 2)) // 40 lines of content into a 10-line viewport
+	// Capture the auto-scrolled-to-bottom position.
+	startOffsetIdx := sl.offsetIdx
+	startOffsetLine := sl.offsetLine
+
+	// User clicks somewhere in the visible area, starting a drag-select.
+	if !sl.HandleMouseDown(5, 3) {
+		t.Fatalf("HandleMouseDown should accept a click inside the viewport")
+	}
+	if !sl.IsMouseDown() {
+		t.Fatalf("IsMouseDown should be true after HandleMouseDown")
+	}
+
+	// New content arrives. With autoScroll still true, SetItems would
+	// normally call GotoBottom() and shift the viewport. The fix should
+	// suppress that while MouseDown is held.
+	sl.SetItems(makeItems(30, 2)) // 60 lines now
+	if sl.offsetIdx != startOffsetIdx || sl.offsetLine != startOffsetLine {
+		t.Errorf("viewport scrolled during active drag: was (%d,%d), now (%d,%d)",
+			startOffsetIdx, startOffsetLine, sl.offsetIdx, sl.offsetLine)
+	}
+
+	// User releases the mouse — drag is over.
+	sl.HandleMouseUp()
+	if sl.IsMouseDown() {
+		t.Fatalf("IsMouseDown should be false after HandleMouseUp")
+	}
+
+	// After release, a fresh content update should resume auto-scrolling
+	// (move the offset to track the new bottom).
+	afterReleaseIdx := sl.offsetIdx
+	afterReleaseLine := sl.offsetLine
+	sl.SetItems(makeItems(50, 2))
+	if sl.offsetIdx == afterReleaseIdx && sl.offsetLine == afterReleaseLine {
+		t.Errorf("autoscroll did not resume after MouseUp: offset stuck at (%d,%d)",
+			afterReleaseIdx, afterReleaseLine)
+	}
+}
+
+// TestScrollList_DragDisablesAutoScroll verifies that any successful
+// HandleMouseDrag call clears autoScroll, even when HandleMouseDown didn't
+// observe it (e.g. a stale wheel-down event set it back to true mid-stream).
+func TestScrollList_DragDisablesAutoScroll(t *testing.T) {
+	sl := NewScrollList(80, 10)
+	sl.SetItems(makeItems(20, 2))
+
+	// Begin a selection.
+	if !sl.HandleMouseDown(5, 3) {
+		t.Fatalf("HandleMouseDown failed")
+	}
+	// Simulate an external code path that re-enabled autoScroll while
+	// MouseDown is still held (the precise condition that caused drift).
+	sl.autoScroll = true
+
+	// Drag motion should hard-lock the viewport again.
+	if !sl.HandleMouseDrag(10, 4) {
+		t.Fatalf("HandleMouseDrag failed")
+	}
+	if sl.autoScroll {
+		t.Errorf("HandleMouseDrag must clear autoScroll to prevent mid-drag jumps")
+	}
+}
+
+// TestScrollList_SetItemsRespectsMouseDown is the most direct regression
+// test: even with autoScroll enabled and new content appended at the
+// bottom, SetItems must not move the viewport while a mouse drag is in
+// progress. This is what caused the "highlighting shifts by 1+ rows
+// during streaming" symptom reported by the user.
+func TestScrollList_SetItemsRespectsMouseDown(t *testing.T) {
+	sl := NewScrollList(80, 5)
+	sl.SetItems(makeItems(10, 2)) // 20 lines into a 5-line viewport
+	// At bottom.
+	preIdx, preLine := sl.offsetIdx, sl.offsetLine
+
+	// Hold mouse down (no actual drag needed).
+	if !sl.HandleMouseDown(0, 0) {
+		t.Fatalf("HandleMouseDown failed")
+	}
+
+	// Append several more items as if streaming. With the bug, each
+	// SetItems would call GotoBottom and shift the offset.
+	for n := 11; n <= 15; n++ {
+		sl.SetItems(makeItems(n, 2))
+		if sl.offsetIdx != preIdx || sl.offsetLine != preLine {
+			t.Fatalf("viewport drifted during streaming with mouse held: "+
+				"start=(%d,%d) now=(%d,%d) after adding item %d",
+				preIdx, preLine, sl.offsetIdx, sl.offsetLine, n)
+		}
+	}
+}

pkg/kit/README.md

@@ -243,7 +243,7 @@ host.ClearSession()
 
 ## Re-exported Types
 
-The SDK re-exports types so you don't need direct internal imports:
+The SDK re-exports message/session/MCP types so you don't need direct internal imports. Agent-configuration types are Kit-owned (not aliases) and use only SDK types in their signatures, so consumers never need to import the underlying LLM-provider package.
 
 ```go
 // Message types
@@ -251,13 +251,28 @@ kit.Message, kit.MessageRole, kit.ContentPart
 kit.TextContent, kit.ReasoningContent, kit.ToolCall, kit.ToolResult, kit.Finish
 kit.RoleUser, kit.RoleAssistant, kit.RoleTool, kit.RoleSystem
 
-// LLM types — concrete Kit-owned structs, no external library dependency
+// LLM types — Kit-owned `LLM*` aliases over the underlying provider types,
+// so consumers never import the provider package directly
 kit.LLMMessage      // {Role LLMMessageRole, Content string}
 kit.LLMMessageRole  // "user" | "assistant" | "system" | "tool"
 kit.LLMUsage        // {InputTokens, OutputTokens, TotalTokens, ...}
 kit.LLMResponse     // {Content, FinishReason, Usage}
 kit.LLMFilePart     // {Filename, Data []byte, MediaType}
 
+// Agent configuration — concrete Kit-owned structs and function types.
+// All fields use SDK types (e.g. `[]kit.Tool`), so consumers can construct
+// these without importing any LLM-provider package.
+kit.AgentConfig              // Lower-level agent config — prefer Options unless you need direct control
+kit.DebugLogger              // Interface: LogDebug(string) / IsDebugEnabled() bool
+kit.MCPTaskConfig            // Task-aware MCP tools/call config (modes, polling, progress)
+kit.ToolCallHandler          // func(toolCallID, toolName, toolArgs string)
+kit.ToolExecutionHandler     // func(toolCallID, toolName, toolArgs string, isStarting bool)
+kit.ToolResultHandler        // func(toolCallID, toolName, toolArgs, result, metadata string, isError bool)
+kit.ResponseHandler          // func(content string)
+kit.StreamingResponseHandler // func(content string)
+kit.ToolCallContentHandler   // func(content string)
+kit.SpinnerFunc              // func(fn func() error) error
+
 // MCP OAuth types
 kit.MCPServer            // *server.MCPServer for in-process MCP transport
 kit.MCPServerConfig      // Configuration for an MCP server (stdio, SSE, or in-process)

pkg/kit/agent_config_internal_test.go

@@ -0,0 +1,208 @@
+package kit
+
+import (
+	"context"
+	"errors"
+	"testing"
+	"time"
+
+	"github.com/mark3labs/kit/internal/agent"
+)
+
+// TestAgentConfigToInternal verifies that the SDK-side AgentConfig converts
+// faithfully to the internal agent.AgentConfig representation, preserving
+// every field consumed by the internal agent layer.
+//
+// Regression test for https://github.com/mark3labs/kit/issues/30.
+func TestAgentConfigToInternal(t *testing.T) {
+	t.Run("nil receiver returns nil", func(t *testing.T) {
+		var c *AgentConfig
+		if got := c.toInternal(); got != nil {
+			t.Errorf("nil.toInternal() = %v, want nil", got)
+		}
+	})
+
+	t.Run("scalar fields round-trip", func(t *testing.T) {
+		c := &AgentConfig{
+			SystemPrompt:     "sys",
+			MaxSteps:         7,
+			StreamingEnabled: true,
+			DisableCoreTools: true,
+		}
+		got := c.toInternal()
+		if got == nil {
+			t.Fatal("toInternal() = nil")
+		}
+		if got.SystemPrompt != "sys" {
+			t.Errorf("SystemPrompt = %q, want %q", got.SystemPrompt, "sys")
+		}
+		if got.MaxSteps != 7 {
+			t.Errorf("MaxSteps = %d, want 7", got.MaxSteps)
+		}
+		if !got.StreamingEnabled {
+			t.Error("StreamingEnabled = false, want true")
+		}
+		if !got.DisableCoreTools {
+			t.Error("DisableCoreTools = false, want true")
+		}
+	})
+
+	t.Run("tool slices propagate without conversion", func(t *testing.T) {
+		// Tool is a type alias for the underlying LLM-tool type, so the
+		// SDK []Tool and internal []fantasy.AgentTool slices share the
+		// same backing array after conversion.
+		tool := NewTool[struct{}]("noop", "noop", nil)
+		c := &AgentConfig{
+			CoreTools:  []Tool{tool},
+			ExtraTools: []Tool{tool, tool},
+		}
+		got := c.toInternal()
+		if len(got.CoreTools) != 1 {
+			t.Errorf("CoreTools len = %d, want 1", len(got.CoreTools))
+		}
+		if len(got.ExtraTools) != 2 {
+			t.Errorf("ExtraTools len = %d, want 2", len(got.ExtraTools))
+		}
+	})
+
+	t.Run("tool wrapper is invoked through internal config", func(t *testing.T) {
+		called := false
+		c := &AgentConfig{
+			ToolWrapper: func(in []Tool) []Tool {
+				called = true
+				return in
+			},
+		}
+		got := c.toInternal()
+		if got.ToolWrapper == nil {
+			t.Fatal("internal ToolWrapper is nil")
+		}
+		_ = got.ToolWrapper(nil)
+		if !called {
+			t.Error("SDK ToolWrapper was not invoked through the internal config")
+		}
+	})
+
+	t.Run("OnMCPServerLoaded propagates", func(t *testing.T) {
+		var captured string
+		wantErr := errors.New("boom")
+		c := &AgentConfig{
+			OnMCPServerLoaded: func(name string, _ int, _ error) {
+				captured = name
+			},
+		}
+		got := c.toInternal()
+		got.OnMCPServerLoaded("svr", 3, wantErr)
+		if captured != "svr" {
+			t.Errorf("OnMCPServerLoaded captured = %q, want %q", captured, "svr")
+		}
+	})
+
+	t.Run("DebugLogger propagates", func(t *testing.T) {
+		dl := &fakeDebugLogger{enabled: true}
+		c := &AgentConfig{DebugLogger: dl}
+		got := c.toInternal()
+		if got.DebugLogger == nil {
+			t.Fatal("internal DebugLogger is nil")
+		}
+		if !got.DebugLogger.IsDebugEnabled() {
+			t.Error("IsDebugEnabled = false, want true")
+		}
+		got.DebugLogger.LogDebug("hello")
+		if len(dl.messages) != 1 || dl.messages[0] != "hello" {
+			t.Errorf("messages = %v, want [hello]", dl.messages)
+		}
+	})
+
+	t.Run("MCPTaskConfig propagates with mode + progress", func(t *testing.T) {
+		c := &AgentConfig{
+			MCPTaskConfig: MCPTaskConfig{
+				PerServerMode: map[string]MCPTaskMode{
+					"build-svr": MCPTaskModeAlways,
+				},
+				DefaultTTL:      30 * time.Second,
+				PollInterval:    250 * time.Millisecond,
+				MaxPollInterval: 2 * time.Second,
+				Timeout:         5 * time.Minute,
+				Progress:        func(_ MCPTaskProgress) {},
+			},
+		}
+		got := c.toInternal()
+		if got.MCPTaskConfig.DefaultTTL != 30*time.Second {
+			t.Errorf("DefaultTTL = %v, want 30s", got.MCPTaskConfig.DefaultTTL)
+		}
+		if got.MCPTaskConfig.PollInterval != 250*time.Millisecond {
+			t.Errorf("PollInterval = %v, want 250ms", got.MCPTaskConfig.PollInterval)
+		}
+		if got.MCPTaskConfig.MaxPollInterval != 2*time.Second {
+			t.Errorf("MaxPollInterval = %v, want 2s", got.MCPTaskConfig.MaxPollInterval)
+		}
+		if got.MCPTaskConfig.Timeout != 5*time.Minute {
+			t.Errorf("Timeout = %v, want 5m", got.MCPTaskConfig.Timeout)
+		}
+		mode, ok := got.MCPTaskConfig.PerServerMode["build-svr"]
+		if !ok {
+			t.Fatal("PerServerMode missing 'build-svr'")
+		}
+		if string(mode) != string(MCPTaskModeAlways) {
+			t.Errorf("mode = %q, want %q", mode, MCPTaskModeAlways)
+		}
+		if got.MCPTaskConfig.Progress == nil {
+			t.Fatal("internal Progress handler is nil")
+		}
+	})
+
+	t.Run("auth and token store factories are wired", func(t *testing.T) {
+		auth := &fakeAuthHandler{}
+		tokenCalls := 0
+		var tokenServer string
+		factory := MCPTokenStoreFactory(func(server string) (MCPTokenStore, error) {
+			tokenCalls++
+			tokenServer = server
+			return nil, nil
+		})
+		c := &AgentConfig{
+			AuthHandler:       auth,
+			TokenStoreFactory: factory,
+		}
+		got := c.toInternal()
+		if got.AuthHandler == nil {
+			t.Fatal("internal AuthHandler is nil")
+		}
+		if got.TokenStoreFactory == nil {
+			t.Fatal("internal TokenStoreFactory is nil")
+		}
+		_, _ = got.TokenStoreFactory("https://example.test")
+		if tokenCalls != 1 {
+			t.Errorf("token factory call count = %d, want 1", tokenCalls)
+		}
+		if tokenServer != "https://example.test" {
+			t.Errorf("token factory server arg = %q", tokenServer)
+		}
+		if got.AuthHandler.RedirectURI() != "redirect" {
+			t.Errorf("RedirectURI = %q, want %q", got.AuthHandler.RedirectURI(), "redirect")
+		}
+	})
+
+	// Compile-time check that the internal type is what we expect.
+	//nolint:staticcheck // QF1011: explicit type asserts the conversion target.
+	var _ *agent.AgentConfig = (&AgentConfig{}).toInternal()
+}
+
+// fakeAuthHandler implements both kit.MCPAuthHandler and the structurally
+// identical tools.MCPAuthHandler used by the internal layer.
+type fakeAuthHandler struct{}
+
+func (f *fakeAuthHandler) RedirectURI() string { return "redirect" }
+func (f *fakeAuthHandler) HandleAuth(_ context.Context, _ string, _ string) (string, error) {
+	return "", nil
+}
+
+// fakeDebugLogger implements kit.DebugLogger for tests.
+type fakeDebugLogger struct {
+	enabled  bool
+	messages []string
+}
+
+func (f *fakeDebugLogger) LogDebug(m string)    { f.messages = append(f.messages, m) }
+func (f *fakeDebugLogger) IsDebugEnabled() bool { return f.enabled }

pkg/kit/extension_api.go

@@ -76,6 +76,22 @@ type ExtensionAPI interface {
 	// Lifecycle
 	Reload() error
 	HasExtensions() bool
+
+	// Loaded returns metadata about the extensions currently loaded.
+	Loaded() []ExtensionInfo
+}
+
+// ExtensionInfo describes a single loaded extension for display purposes
+// (e.g. the startup banner or `kit extensions list`).
+type ExtensionInfo struct {
+	// Path is the absolute path of the extension's .go file.
+	Path string
+	// ToolCount is the number of tools registered by the extension.
+	ToolCount int
+	// CommandCount is the number of slash commands registered.
+	CommandCount int
+	// HandlerCount is the total number of event handlers registered.
+	HandlerCount int
 }
 
 // extensionAPI implements ExtensionAPI by wrapping a Kit instance.
@@ -456,3 +472,27 @@ func (e *extensionAPI) Reload() error {
 func (e *extensionAPI) HasExtensions() bool {
 	return e.kit.extRunner != nil
 }
+
+func (e *extensionAPI) Loaded() []ExtensionInfo {
+	if e.kit.extRunner == nil {
+		return nil
+	}
+	exts := e.kit.extRunner.Extensions()
+	if len(exts) == 0 {
+		return nil
+	}
+	infos := make([]ExtensionInfo, 0, len(exts))
+	for _, ex := range exts {
+		handlerCount := 0
+		for _, hs := range ex.Handlers {
+			handlerCount += len(hs)
+		}
+		infos = append(infos, ExtensionInfo{
+			Path:         ex.Path,
+			ToolCount:    len(ex.Tools),
+			CommandCount: len(ex.Commands),
+			HandlerCount: handlerCount,
+		})
+	}
+	return infos
+}

pkg/kit/kit.go

@@ -1489,7 +1489,7 @@ func New(ctx context.Context, opts *Options) (*Kit, error) {
 
 	if opts.CLI != nil {
 		setupOpts.ShowSpinner = opts.CLI.ShowSpinner
-		setupOpts.SpinnerFunc = opts.CLI.SpinnerFunc
+		setupOpts.SpinnerFunc = agent.SpinnerFunc(opts.CLI.SpinnerFunc)
 		setupOpts.UseBufferedLogger = opts.CLI.UseBufferedLogger
 		if opts.CLI.ProgressReaderFunc != nil {
 			providerConfig.ProgressReaderFunc = opts.CLI.ProgressReaderFunc

pkg/kit/mcp_tasks.go

@@ -98,6 +98,70 @@ type MCPTaskProgress struct {
 // dispatched on a goroutine.
 type MCPTaskProgressHandler func(MCPTaskProgress)
 
+// MCPTaskConfig configures task-aware MCP tools/call execution. All fields
+// are optional; the zero value disables progress callbacks and applies
+// sensible polling defaults inside the engine.
+//
+// For most consumers, the flat [Options] fields (`MCPTaskMode`,
+// `MCPTaskTTL`, `MCPTaskPollInterval`, `MCPTaskMaxPollInterval`,
+// `MCPTaskTimeout`, `MCPTaskProgress`) are the preferred entry point.
+// MCPTaskConfig is exposed for the low-level [AgentConfig] path.
+type MCPTaskConfig struct {
+	// PerServerMode overrides the per-server task mode resolved from
+	// [MCPServerConfig]. Keys are server names. Missing entries fall back
+	// to the configured value.
+	PerServerMode map[string]MCPTaskMode
+
+	// DefaultTTL is the TTL hint sent in TaskParams when augmenting a
+	// tools/call. Zero means omit the TTL — let the server pick its own.
+	DefaultTTL time.Duration
+
+	// PollInterval is the fallback interval between tasks/get requests
+	// when the server does not suggest one. Zero defaults to 1 second.
+	PollInterval time.Duration
+
+	// MaxPollInterval caps the polling interval. Zero defaults to 5 seconds.
+	MaxPollInterval time.Duration
+
+	// Timeout is the maximum wall-clock duration to wait for a task to
+	// reach a terminal state. Zero defaults to 15 minutes. Independent
+	// of the per-call context deadline; whichever fires first wins.
+	Timeout time.Duration
+
+	// Progress, if non-nil, receives every status transition observed by
+	// the polling loop.
+	Progress MCPTaskProgressHandler
+}
+
+// toToolsConfig converts the SDK-level [MCPTaskConfig] to the internal
+// tools-package representation. Keeps the dependency arrow internal-only.
+func (c MCPTaskConfig) toToolsConfig() tools.MCPTaskConfig {
+	cfg := tools.MCPTaskConfig{
+		DefaultTTL:      c.DefaultTTL,
+		PollInterval:    c.PollInterval,
+		MaxPollInterval: c.MaxPollInterval,
+		Timeout:         c.Timeout,
+	}
+	if len(c.PerServerMode) > 0 {
+		cfg.PerServerMode = make(map[string]tools.MCPTaskMode, len(c.PerServerMode))
+		for k, v := range c.PerServerMode {
+			cfg.PerServerMode[k] = tools.MCPTaskMode(v)
+		}
+	}
+	if c.Progress != nil {
+		h := c.Progress
+		cfg.Progress = func(p tools.MCPTaskProgress) {
+			h(MCPTaskProgress{
+				Server:  p.Server,
+				TaskID:  p.TaskID,
+				Status:  MCPTaskStatus(p.Status),
+				Message: p.Message,
+			})
+		}
+	}
+	return cfg
+}
+
 // mcpTaskOptions carries SDK consumer configuration into the agent setup.
 // Stored on Options as a single value so the public surface stays compact;
 // individual fields are exposed via WithMCP* builder functions.

pkg/kit/types.go

@@ -11,6 +11,7 @@ import (
 	"github.com/mark3labs/kit/internal/message"
 	"github.com/mark3labs/kit/internal/models"
 	"github.com/mark3labs/kit/internal/session"
+	"github.com/mark3labs/kit/internal/tools"
 	"github.com/mark3labs/mcp-go/client/transport"
 	"github.com/mark3labs/mcp-go/server"
 )
@@ -75,25 +76,151 @@ type Config = config.Config
 // local (stdio) and remote (StreamableHTTP/SSE) server types.
 type MCPServerConfig = config.MCPServerConfig
 
-// ==== Agent Types (internal/agent/) ====
+// ==== Agent Types ====
 
-// AgentConfig holds configuration options for creating a new Agent.
-type AgentConfig = agent.AgentConfig
+// DebugLogger is an SDK-owned interface for low-level debug logging from
+// the engine and MCP tool plumbing. Implementations must be safe for
+// concurrent use.
+//
+// Most consumers do not need to provide one; pass [Options.Debug] = true
+// to use the default logger. DebugLogger is exposed for the low-level
+// [AgentConfig] path and for embedders that want to route debug output
+// into their own logging system.
+type DebugLogger interface {
+	// LogDebug records a single debug message. Implementations may drop,
+	// buffer, or render the message however they choose.
+	LogDebug(message string)
+	// IsDebugEnabled reports whether debug logging is active. Callers may
+	// check this before doing expensive formatting work.
+	IsDebugEnabled() bool
+}
 
-type (
-	// ToolCallHandler is a function type for handling tool calls as they happen.
-	ToolCallHandler = agent.ToolCallHandler
-	// ToolExecutionHandler is a function type for handling tool execution start/end events.
-	ToolExecutionHandler = agent.ToolExecutionHandler
-	// ToolResultHandler is a function type for handling tool results.
-	ToolResultHandler = agent.ToolResultHandler
-	// ResponseHandler is a function type for handling LLM responses.
-	ResponseHandler = agent.ResponseHandler
-	// StreamingResponseHandler is a function type for handling streaming LLM responses.
-	StreamingResponseHandler = agent.StreamingResponseHandler
-	// ToolCallContentHandler is a function type for handling content that accompanies tool calls.
-	ToolCallContentHandler = agent.ToolCallContentHandler
-)
+// AgentConfig holds configuration options for constructing an agent at the
+// SDK boundary. All fields use SDK-owned types, so consumers can populate
+// this struct without importing any underlying LLM-provider package.
+//
+// For most use cases, prefer the high-level [New] entry point with
+// [Options]. AgentConfig is exposed for advanced consumers that need
+// direct access to the lower-level agent configuration shape.
+type AgentConfig struct {
+	// ModelConfig holds the LLM provider configuration. A nil value means
+	// that the default provider/model resolution will be used.
+	ModelConfig *ProviderConfig
+
+	// MCPConfig describes any MCP servers whose tools should be loaded
+	// alongside core tools.
+	MCPConfig *Config
+
+	// SystemPrompt is the system prompt sent to the LLM.
+	SystemPrompt string
+
+	// MaxSteps caps the number of LLM iterations per turn. A value of
+	// zero means no cap is applied at this layer.
+	MaxSteps int
+
+	// StreamingEnabled controls whether the agent streams responses.
+	StreamingEnabled bool
+
+	// AuthHandler handles OAuth authorization for remote MCP servers.
+	// When nil, remote MCP servers requiring OAuth will fail to connect.
+	AuthHandler MCPAuthHandler
+
+	// TokenStoreFactory, if non-nil, creates a custom token store for each
+	// remote MCP server's OAuth tokens. When nil, the default file-based
+	// token store is used.
+	TokenStoreFactory MCPTokenStoreFactory
+
+	// CoreTools overrides the default core tool set. If empty, [AllTools]
+	// is used. Provide a custom tool set (e.g. [CodingTools] or tools
+	// built with a custom WorkDir) to scope agent capabilities.
+	CoreTools []Tool
+
+	// DisableCoreTools, when true, prevents loading any core tools.
+	// Combined with empty CoreTools this yields a chat-only agent with
+	// no built-in tools.
+	DisableCoreTools bool
+
+	// ExtraTools are additional tools loaded alongside core and MCP tools.
+	ExtraTools []Tool
+
+	// ToolWrapper, if non-nil, wraps the combined tool list before it is
+	// handed to the LLM. Used to intercept tool calls or results.
+	ToolWrapper func([]Tool) []Tool
+
+	// OnMCPServerLoaded, if non-nil, is invoked once for each MCP server
+	// when its tools have finished loading (or failed). Called from a
+	// background goroutine.
+	OnMCPServerLoaded func(serverName string, toolCount int, err error)
+
+	// DebugLogger receives low-level debug output from the engine and the
+	// MCP tool plumbing. Nil means no debug output is emitted at this
+	// layer (regardless of [Options.Debug], which feeds the higher-level
+	// [New] entry point). Pass an implementation here when wiring a custom
+	// logger through the lower-level AgentConfig path.
+	DebugLogger DebugLogger
+
+	// MCPTaskConfig configures task-aware MCP tools/call execution — mode
+	// overrides, polling intervals, timeouts, and the progress handler.
+	// The zero value preserves historical synchronous-only behaviour for
+	// any server that didn't advertise task support during initialize.
+	MCPTaskConfig MCPTaskConfig
+}
+
+// toInternal converts an AgentConfig to its internal representation.
+// Slice and function fields convert without allocation because [Tool]
+// is a type alias for the underlying LLM-tool type.
+func (c *AgentConfig) toInternal() *agent.AgentConfig {
+	if c == nil {
+		return nil
+	}
+	out := &agent.AgentConfig{
+		ModelConfig:       c.ModelConfig,
+		MCPConfig:         c.MCPConfig,
+		SystemPrompt:      c.SystemPrompt,
+		MaxSteps:          c.MaxSteps,
+		StreamingEnabled:  c.StreamingEnabled,
+		CoreTools:         c.CoreTools,
+		DisableCoreTools:  c.DisableCoreTools,
+		ExtraTools:        c.ExtraTools,
+		ToolWrapper:       c.ToolWrapper,
+		OnMCPServerLoaded: c.OnMCPServerLoaded,
+	}
+	if c.AuthHandler != nil {
+		out.AuthHandler = c.AuthHandler
+	}
+	if c.TokenStoreFactory != nil {
+		out.TokenStoreFactory = tools.TokenStoreFactory(c.TokenStoreFactory)
+	}
+	if c.DebugLogger != nil {
+		out.DebugLogger = c.DebugLogger
+	}
+	out.MCPTaskConfig = c.MCPTaskConfig.toToolsConfig()
+	return out
+}
+
+// ToolCallHandler is invoked when the LLM produces a tool call. It receives
+// the call ID, tool name, and the JSON-encoded input arguments.
+type ToolCallHandler func(toolCallID, toolName, toolArgs string)
+
+// ToolExecutionHandler is invoked at the start and end of tool execution.
+// The isStarting flag distinguishes the two phases.
+type ToolExecutionHandler func(toolCallID, toolName, toolArgs string, isStarting bool)
+
+// ToolResultHandler is invoked after a tool finishes executing. The metadata
+// parameter carries optional structured data (e.g. file-diff info) from the
+// tool execution, JSON-encoded; it may be empty.
+type ToolResultHandler func(toolCallID, toolName, toolArgs, result, metadata string, isError bool)
+
+// ResponseHandler is invoked with the final assistant text for each turn.
+type ResponseHandler func(content string)
+
+// StreamingResponseHandler is invoked with each streamed text delta as it
+// arrives from the LLM.
+type StreamingResponseHandler func(content string)
+
+// ToolCallContentHandler is invoked with any assistant text that accompanies
+// a tool call within the same step.
+type ToolCallContentHandler func(content string)
 
 // ==== Provider & Model Types (internal/models/) ====
 
@@ -126,7 +253,7 @@ type ModelsRegistry = models.ModelsRegistry
 
 // SpinnerFunc wraps a function in a loading spinner animation. Used for
 // Ollama model loading. Signature: func(fn func() error) error.
-type SpinnerFunc = agent.SpinnerFunc
+type SpinnerFunc func(fn func() error) error
 
 // ==== LLM Types ====
 //

pkg/kit/types_test.go

@@ -1,6 +1,7 @@
 package kit_test
 
 import (
+	"context"
 	"encoding/json"
 	"testing"
 
@@ -263,6 +264,101 @@ func TestConvertFromLLMMessage(t *testing.T) {
 	}
 }
 
+// TestAgentConfigNoFantasyImport verifies AgentConfig can be populated with
+// every field — including CoreTools, ExtraTools, and ToolWrapper — using
+// only SDK-owned types. This test deliberately does not import
+// "charm.land/fantasy"; the package compiling at all is the proof that the
+// SDK no longer leaks the dependency name through AgentConfig.
+//
+// Regression test for https://github.com/mark3labs/kit/issues/30.
+func TestAgentConfigNoFantasyImport(t *testing.T) {
+	myTool := kit.NewTool[struct{}]("noop", "does nothing", func(_ context.Context, _ struct{}) (kit.ToolOutput, error) {
+		return kit.TextResult("ok"), nil
+	})
+
+	wrapperCalled := false
+	cfg := kit.AgentConfig{
+		SystemPrompt:     "you are a tester",
+		MaxSteps:         5,
+		StreamingEnabled: true,
+		CoreTools:        []kit.Tool{myTool},
+		ExtraTools:       []kit.Tool{myTool},
+		DisableCoreTools: false,
+		ToolWrapper: func(in []kit.Tool) []kit.Tool {
+			wrapperCalled = true
+			return in
+		},
+		OnMCPServerLoaded: func(_ string, _ int, _ error) {},
+	}
+
+	if cfg.SystemPrompt != "you are a tester" {
+		t.Errorf("SystemPrompt = %q, want %q", cfg.SystemPrompt, "you are a tester")
+	}
+	if cfg.MaxSteps != 5 {
+		t.Errorf("MaxSteps = %d, want 5", cfg.MaxSteps)
+	}
+	if !cfg.StreamingEnabled {
+		t.Error("StreamingEnabled = false, want true")
+	}
+	if len(cfg.CoreTools) != 1 {
+		t.Errorf("CoreTools len = %d, want 1", len(cfg.CoreTools))
+	}
+	if len(cfg.ExtraTools) != 1 {
+		t.Errorf("ExtraTools len = %d, want 1", len(cfg.ExtraTools))
+	}
+
+	// Exercise the wrapper to confirm the func type is usable.
+	out := cfg.ToolWrapper(cfg.CoreTools)
+	if !wrapperCalled {
+		t.Error("ToolWrapper was not invoked")
+	}
+	if len(out) != 1 {
+		t.Errorf("wrapped tool list len = %d, want 1", len(out))
+	}
+}
+
+// TestAgentConfigToolWrapperSignature documents that AgentConfig.ToolWrapper
+// uses kit.Tool (not the underlying provider type) in its signature.
+func TestAgentConfigToolWrapperSignature(t *testing.T) {
+	//nolint:staticcheck // QF1011: explicit type asserts the SDK-side func signature.
+	var _ func([]kit.Tool) []kit.Tool = func(in []kit.Tool) []kit.Tool { return in }
+	cfg := kit.AgentConfig{
+		ToolWrapper: func(in []kit.Tool) []kit.Tool { return in },
+	}
+	if cfg.ToolWrapper == nil {
+		t.Fatal("ToolWrapper assignment failed")
+	}
+}
+
+// TestSpinnerFuncSignature verifies SpinnerFunc has the documented signature
+// and can be constructed without importing any provider package.
+func TestSpinnerFuncSignature(t *testing.T) {
+	called := false
+	var sp kit.SpinnerFunc = func(fn func() error) error {
+		called = true
+		return fn()
+	}
+	err := sp(func() error { return nil })
+	if err != nil {
+		t.Errorf("SpinnerFunc returned err: %v", err)
+	}
+	if !called {
+		t.Error("SpinnerFunc did not invoke fn")
+	}
+}
+
+// TestHandlerTypesSignatures verifies the SDK-owned handler function types
+// can be assigned from plain function literals using only standard library
+// types in their signatures (no provider-package import required).
+func TestHandlerTypesSignatures(t *testing.T) {
+	var _ kit.ToolCallHandler = func(_, _, _ string) {}
+	var _ kit.ToolExecutionHandler = func(_, _, _ string, _ bool) {}
+	var _ kit.ToolResultHandler = func(_, _, _, _, _ string, _ bool) {}
+	var _ kit.ResponseHandler = func(_ string) {}
+	var _ kit.StreamingResponseHandler = func(_ string) {}
+	var _ kit.ToolCallContentHandler = func(_ string) {}
+}
+
 // containsStr is a tiny helper to avoid importing strings in test.
 func containsStr(s, substr string) bool {
 	return len(s) >= len(substr) && (s == substr || len(s) > 0 && indexStr(s, substr) >= 0)

specs/README.md

@@ -1,5 +0,0 @@
-# Specs
-
-| Spec | Status | Description |
-|------|--------|-------------|
-| [unified-bubbletea-architecture](unified-bubbletea-architecture.md) | Draft | Replace micro-program pattern with single Bubble Tea program + thick app layer |