[#368] fix-broken-issue-title-session (#371)

* takt: fix-broken-issue-title-session

* takt: fix-broken-issue-title-session

src/__tests__/conversationLoop-resume.test.ts

@@ -226,3 +226,38 @@ describe('/resume command', () => {
     expect(result.action).toBe('cancel');
   });
 });
+
+// =================================================================
+// /go command: summary AI session isolation
+// =================================================================
+describe('/go command', () => {
+  it('should pass sessionId as undefined to summary AI even when conversation has an active session', async () => {
+    // Given: send message (AI responds with sessionId) → /go triggers summary
+    setupRawStdin(toRawInputs(['hello', '/go']));
+
+    const { provider, capture } = createScenarioProvider([
+      // Call 0: user message → AI responds and sets sessionId
+      { content: 'AI response', sessionId: 'session-abc' },
+      // Call 1: /go summary → should NOT inherit sessionId
+      { content: '## Fix broken title\nDetails here' },
+    ]);
+
+    const ctx: SessionContext = {
+      provider: provider as SessionContext['provider'],
+      providerType: 'mock' as SessionContext['providerType'],
+      model: undefined,
+      lang: 'en',
+      personaName: 'interactive',
+      sessionId: undefined,
+    };
+
+    // When
+    const result = await runConversationLoop('/test', ctx, defaultStrategy, undefined, undefined);
+
+    // Then: first AI call had no session (initial state)
+    expect(capture.sessionIds[0]).toBeUndefined();
+    // Then: summary call must NOT inherit the conversation session
+    expect(capture.sessionIds[1]).toBeUndefined();
+    expect(result.action).toBe('execute');
+  });
+});

src/__tests__/createIssueFromTask.test.ts

@@ -34,6 +34,7 @@ vi.mock('../shared/utils/index.js', async (importOriginal) => ({
 
 import { success, error } from '../shared/ui/index.js';
 import { createIssueFromTask } from '../features/tasks/index.js';
+import { extractTitle } from '../features/tasks/add/index.js';
 
 const mockSuccess = vi.mocked(success);
 const mockError = vi.mocked(error);
@@ -229,3 +230,149 @@ describe('createIssueFromTask', () => {
     });
   });
 });
+
+describe('extractTitle', () => {
+  describe('Markdown heading extraction', () => {
+    it('should strip # and return text for h1 heading', () => {
+      // Given
+      const task = '# Fix broken title\nDetails here';
+
+      // When
+      const result = extractTitle(task);
+
+      // Then
+      expect(result).toBe('Fix broken title');
+    });
+
+    it('should strip ## and return text for h2 heading', () => {
+      // Given
+      const task = '## Fix broken title\nDetails here';
+
+      // When
+      const result = extractTitle(task);
+
+      // Then
+      expect(result).toBe('Fix broken title');
+    });
+
+    it('should strip ### and return text for h3 heading', () => {
+      // Given
+      const task = '### Fix broken title\nDetails here';
+
+      // When
+      const result = extractTitle(task);
+
+      // Then
+      expect(result).toBe('Fix broken title');
+    });
+
+    it('should prefer first Markdown heading over preceding plain text', () => {
+      // Given: AI preamble followed by heading
+      const task = '失礼しました。修正します。\n## Fix broken title\nDetails here';
+
+      // When
+      const result = extractTitle(task);
+
+      // Then: heading wins over first line
+      expect(result).toBe('Fix broken title');
+    });
+
+    it('should find heading even when multiple empty lines precede it', () => {
+      // Given
+      const task = '\n\n## Fix broken title\nDetails here';
+
+      // When
+      const result = extractTitle(task);
+
+      // Then
+      expect(result).toBe('Fix broken title');
+    });
+  });
+
+  describe('fallback to first non-empty line', () => {
+    it('should return first non-empty line when no Markdown heading exists', () => {
+      // Given: plain text without heading
+      const task = 'Fix broken title\nSecond line details';
+
+      // When
+      const result = extractTitle(task);
+
+      // Then
+      expect(result).toBe('Fix broken title');
+    });
+
+    it('should skip leading empty lines when no heading exists', () => {
+      // Given: leading blank lines
+      const task = '\n\nFix broken title\nDetails here';
+
+      // When
+      const result = extractTitle(task);
+
+      // Then
+      expect(result).toBe('Fix broken title');
+    });
+
+    it('should not treat h4+ headings as Markdown headings', () => {
+      // Given: #### is not matched (only h1-h3 are recognized)
+      const task = '#### h4 heading\nDetails here';
+
+      // When
+      const result = extractTitle(task);
+
+      // Then: falls back to first non-empty line as-is
+      expect(result).toBe('#### h4 heading');
+    });
+
+    it('should not treat heading without space after hash as Markdown heading', () => {
+      // Given: #Title has no space, so not recognized as heading
+      const task = '#NoSpace\nDetails here';
+
+      // When
+      const result = extractTitle(task);
+
+      // Then: falls back to first non-empty line
+      expect(result).toBe('#NoSpace');
+    });
+  });
+
+  describe('title truncation', () => {
+    it('should truncate heading title to 97 chars + ellipsis when over 100 chars', () => {
+      // Given: heading text over 100 characters
+      const longTitle = 'a'.repeat(102);
+      const task = `## ${longTitle}\nDetails here`;
+
+      // When
+      const result = extractTitle(task);
+
+      // Then: truncated to 97 + "..."
+      expect(result).toBe(`${'a'.repeat(97)}...`);
+      expect(result).toHaveLength(100);
+    });
+
+    it('should truncate plain text title to 97 chars + ellipsis when over 100 chars', () => {
+      // Given: plain text over 100 characters
+      const longTitle = 'b'.repeat(102);
+      const task = `${longTitle}\nDetails here`;
+
+      // When
+      const result = extractTitle(task);
+
+      // Then: truncated to 97 + "..."
+      expect(result).toBe(`${'b'.repeat(97)}...`);
+      expect(result).toHaveLength(100);
+    });
+
+    it('should not truncate title of exactly 100 characters', () => {
+      // Given: title exactly 100 chars
+      const title100 = 'c'.repeat(100);
+      const task = `## ${title100}\nDetails here`;
+
+      // When
+      const result = extractTitle(task);
+
+      // Then: not truncated
+      expect(result).toBe(title100);
+      expect(result).toHaveLength(100);
+    });
+  });
+});

src/__tests__/quietMode-session.test.ts

@@ -0,0 +1,160 @@
+/**
+ * Tests for quietMode summary AI session isolation.
+ *
+ * Verifies that the summary AI call in quietMode does not inherit the
+ * conversation session (sessionId must be undefined), even when ctx
+ * carries an active sessionId. This matches the fix already applied to
+ * conversationLoop.ts's /go command.
+ */
+
+import { describe, it, expect, vi, beforeEach } from 'vitest';
+
+// ── Mocks ─────────────────────────────────────────────────────────────
+
+vi.mock('../features/interactive/conversationLoop.js', () => ({
+  initializeSession: vi.fn(),
+  callAIWithRetry: vi.fn(),
+}));
+
+vi.mock('../features/interactive/interactive.js', () => ({
+  DEFAULT_INTERACTIVE_TOOLS: ['Read'],
+  buildSummaryPrompt: vi.fn(),
+  selectPostSummaryAction: vi.fn(),
+}));
+
+vi.mock('../shared/utils/index.js', async (importOriginal) => ({
+  ...(await importOriginal<Record<string, unknown>>()),
+  createLogger: () => ({ info: vi.fn(), debug: vi.fn(), error: vi.fn() }),
+}));
+
+vi.mock('../shared/ui/index.js', () => ({
+  info: vi.fn(),
+  error: vi.fn(),
+  blankLine: vi.fn(),
+}));
+
+vi.mock('../shared/i18n/index.js', () => ({
+  getLabel: vi.fn((_key: string, _lang: string) => 'Mock label'),
+  getLabelObject: vi.fn(() => ({
+    intro: 'Intro',
+    proposed: 'Proposed:',
+    cancelled: 'Cancelled',
+  })),
+}));
+
+// ── Imports (after mocks) ──────────────────────────────────────────────
+
+import { quietMode } from '../features/interactive/quietMode.js';
+import { initializeSession, callAIWithRetry } from '../features/interactive/conversationLoop.js';
+import { buildSummaryPrompt, selectPostSummaryAction } from '../features/interactive/interactive.js';
+import type { SessionContext } from '../features/interactive/conversationLoop.js';
+
+const mockInitializeSession = vi.mocked(initializeSession);
+const mockCallAIWithRetry = vi.mocked(callAIWithRetry);
+const mockBuildSummaryPrompt = vi.mocked(buildSummaryPrompt);
+const mockSelectPostSummaryAction = vi.mocked(selectPostSummaryAction);
+
+// ── Helpers ───────────────────────────────────────────────────────────
+
+function createMockSessionContext(sessionId: string | undefined): SessionContext {
+  return {
+    provider: {} as SessionContext['provider'],
+    providerType: 'mock' as SessionContext['providerType'],
+    model: undefined,
+    lang: 'en',
+    personaName: 'interactive',
+    sessionId,
+  };
+}
+
+// ── Setup ─────────────────────────────────────────────────────────────
+
+beforeEach(() => {
+  vi.clearAllMocks();
+});
+
+// =================================================================
+// quietMode: summary AI session isolation
+// =================================================================
+
+describe('quietMode: summary AI session isolation', () => {
+  it('should pass sessionId as undefined to callAIWithRetry even when ctx carries an active sessionId', async () => {
+    // Given: initializeSession returns a ctx with an active session
+    const ctxWithSession = createMockSessionContext('active-session-123');
+    mockInitializeSession.mockReturnValue(ctxWithSession);
+
+    // Given: buildSummaryPrompt returns a non-null prompt
+    mockBuildSummaryPrompt.mockReturnValue('Summary prompt for task');
+
+    // Given: callAIWithRetry returns a successful result
+    mockCallAIWithRetry.mockResolvedValue({
+      result: { content: '## Fix the bug\nDetails here', success: true, sessionId: undefined },
+      sessionId: undefined,
+    });
+
+    // Given: user selects execute action
+    mockSelectPostSummaryAction.mockResolvedValue('execute');
+
+    // When
+    const result = await quietMode('/test/cwd', 'fix the bug');
+
+    // Then: callAIWithRetry was called exactly once
+    expect(mockCallAIWithRetry).toHaveBeenCalledOnce();
+
+    // Then: 5th argument (ctx) must NOT inherit the conversation sessionId
+    const calledCtx = mockCallAIWithRetry.mock.calls[0]![4] as SessionContext;
+    expect(calledCtx.sessionId).toBeUndefined();
+
+    // Then: result is as expected
+    expect(result.action).toBe('execute');
+  });
+
+  it('should preserve other ctx fields while clearing sessionId', async () => {
+    // Given: ctx with active session and specific lang
+    const ctxWithSession = createMockSessionContext('session-xyz');
+    ctxWithSession.lang = 'ja';
+    mockInitializeSession.mockReturnValue(ctxWithSession);
+    mockBuildSummaryPrompt.mockReturnValue('要約プロンプト');
+    mockCallAIWithRetry.mockResolvedValue({
+      result: { content: '## タスク\n詳細', success: true, sessionId: undefined },
+      sessionId: undefined,
+    });
+    mockSelectPostSummaryAction.mockResolvedValue('execute');
+
+    // When
+    await quietMode('/test/cwd', 'バグを修正する');
+
+    // Then: sessionId is cleared but other fields are preserved
+    const calledCtx = mockCallAIWithRetry.mock.calls[0]![4] as SessionContext;
+    expect(calledCtx.sessionId).toBeUndefined();
+    expect(calledCtx.lang).toBe('ja');
+    expect(calledCtx.personaName).toBe('interactive');
+  });
+
+  it('should return cancel when callAIWithRetry returns null result', async () => {
+    // Given
+    mockInitializeSession.mockReturnValue(createMockSessionContext('session-abc'));
+    mockBuildSummaryPrompt.mockReturnValue('Summary prompt');
+    mockCallAIWithRetry.mockResolvedValue({ result: null, sessionId: undefined });
+
+    // When
+    const result = await quietMode('/test/cwd', 'some input');
+
+    // Then
+    expect(result.action).toBe('cancel');
+    expect(result.task).toBe('');
+  });
+
+  it('should return cancel when buildSummaryPrompt returns null', async () => {
+    // Given: no conversation history leads to null prompt
+    mockInitializeSession.mockReturnValue(createMockSessionContext(undefined));
+    mockBuildSummaryPrompt.mockReturnValue(null);
+
+    // When
+    const result = await quietMode('/test/cwd', 'some input');
+
+    // Then: short-circuits before callAIWithRetry
+    expect(mockCallAIWithRetry).not.toHaveBeenCalled();
+    expect(result.action).toBe('cancel');
+  });
+});

src/features/interactive/conversationLoop.ts

@@ -192,7 +192,11 @@ export async function runConversationLoop(
       if (userNote) {
         summaryPrompt = `${summaryPrompt}\n\nUser Note:\n${userNote}`;
       }
-      const summaryResult = await doCallAI(summaryPrompt, summaryPrompt, strategy.allowedTools);
+      // Summary AI must not inherit the conversation session to avoid chat-mode behavior.
+      const { result: summaryResult } = await callAIWithRetry(
+        summaryPrompt, summaryPrompt, strategy.allowedTools, cwd,
+        { ...ctx, sessionId: undefined },
+      );
       if (!summaryResult) {
         info(ui.summarizeFailed);
         continue;

src/features/interactive/quietMode.ts

@@ -85,7 +85,8 @@ export async function quietMode(
   }
 
   const { result } = await callAIWithRetry(
-    summaryPrompt, summaryPrompt, DEFAULT_INTERACTIVE_TOOLS, cwd, ctx,
+    summaryPrompt, summaryPrompt, DEFAULT_INTERACTIVE_TOOLS, cwd,
+    { ...ctx, sessionId: undefined },
   );
 
   if (!result) {

src/features/tasks/add/index.ts

@@ -70,16 +70,38 @@ export async function saveTaskFile(
   return { taskName: created.name, tasksFile };
 }
 
+const TITLE_MAX_LENGTH = 100;
+const TITLE_TRUNCATE_LENGTH = 97;
+const MARKDOWN_HEADING_PATTERN = /^#{1,3}\s+\S/;
+
+/**
+ * Extract a clean title from a task description.
+ *
+ * Prefers the first Markdown heading (h1-h3) if present.
+ * Falls back to the first non-empty line otherwise.
+ * Truncates to 100 characters (97 + "...") when exceeded.
+ */
+export function extractTitle(task: string): string {
+  const lines = task.split('\n');
+  const headingLine = lines.find((l) => MARKDOWN_HEADING_PATTERN.test(l));
+  const titleLine = headingLine
+    ? headingLine.replace(/^#{1,3}\s+/, '')
+    : (lines.find((l) => l.trim().length > 0) ?? task);
+  return titleLine.length > TITLE_MAX_LENGTH
+    ? `${titleLine.slice(0, TITLE_TRUNCATE_LENGTH)}...`
+    : titleLine;
+}
+
 /**
  * Create a GitHub Issue from a task description.
  *
- * Extracts the first line as the issue title (truncated to 100 chars),
- * uses the full task as the body, and displays success/error messages.
+ * Extracts the first Markdown heading (h1-h3) as the issue title,
+ * falling back to the first non-empty line. Truncates to 100 chars.
+ * Uses the full task as the body, and displays success/error messages.
  */
 export function createIssueFromTask(task: string, options?: { labels?: string[] }): number | undefined {
   info('Creating GitHub Issue...');
-  const titleLine = task.split('\n')[0] || task;
-  const title = titleLine.length > 100 ? `${titleLine.slice(0, 97)}...` : titleLine;
+  const title = extractTitle(task);
   const effectiveLabels = options?.labels?.filter((l) => l.length > 0) ?? [];
   const labels = effectiveLabels.length > 0 ? effectiveLabels : undefined;
 
@@ -259,7 +281,6 @@ export async function addTask(cwd: string, task?: string): Promise<void> {
     return;
   }
 
-  // 3. ワークツリー/ブランチ/PR設定
   const settings = await promptWorktreeSettings();
 
   // YAMLファイル作成