fix(dream): stamp incremental extraction watermark (#2636)

The Dream cycle disables sync's inline extraction and routes changed
slugs through extractForSlugs, which flushed link/timeline batches but
never stamped links_extracted_at — so incrementally extracted pages
stayed permanently visible to `extract --stale` / doctor.

Collect processedRefs per successfully processed page and stamp them
via stampExtracted (best-effort) after both batch flushes, non-dry-run
mode 'all' only. Source-id threading from the original PR #2637 already
landed on master via #1503/#1747, so this rebase carries only the
missing watermark stamp plus regression tests.

Takeover of #2637.

Co-authored-by: JavanC <JavanC@users.noreply.github.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

src/commands/extract.ts

@@ -1025,6 +1025,10 @@ async function extractForSlugs(
   let linksCreated = 0;
   let timelineCreated = 0;
   let pagesProcessed = 0;
+  // #2636: successfully processed pages get their extraction watermark
+  // stamped after the final flush (mode 'all' only — a partial-mode run
+  // hasn't done the full extraction the watermark asserts).
+  const processedRefs: Array<{ slug: string; source_id: string }> = [];
 
   // Issue #972: read the basename flag once per extract run.
   const globalBasename = await isGlobalBasenameEnabled(engine);
@@ -1113,6 +1117,7 @@ async function extractForSlugs(
         }
 
         pagesProcessed++;
+        if (!dryRun) processedRefs.push({ slug, source_id: sourceId ?? 'default' });
       } catch { /* skip unreadable */ }
       progress.tick(1);
     },
@@ -1120,6 +1125,13 @@ async function extractForSlugs(
 
   await flushLinks();
   await flushTimeline();
+  // #2636: the Dream cycle disables sync's inline extraction and routes
+  // changed slugs through this incremental path — without a stamp here,
+  // those pages never get links_extracted_at and stay permanently visible
+  // to `extract --stale` / doctor. Stamp only after BOTH batches flushed.
+  if (!dryRun && mode === 'all') {
+    await stampExtracted(engine, processedRefs);
+  }
   progress.finish();
 
   if (!jsonMode) {

src/core/context-engine.ts

@@ -453,14 +453,7 @@ function resolveActivity(
  * every `assemble()` call. 1 MB is generous for a human-edited task list. */
 const MAX_TASKS_MD_BYTES = 1_000_000;
 
-/** Extract open tasks from ops/tasks.md Today section.
- *
- * The daily-task-manager skill's documented Output Format uses priority
- * headings (`## P1 — Today`) with plain `- [ ] task` lines; older fixtures
- * used a bare `## Today` heading with bold task names. Accept both so the
- * live-context reader matches the documented writer contract instead of
- * silently surfacing no tasks (#2186).
- */
+/** Extract open tasks from ops/tasks.md "## Today" section. */
 function resolveTodayTasks(workspaceDir: string): string[] {
   try {
     const path = join(workspaceDir, 'ops', 'tasks.md');
@@ -468,18 +461,14 @@ function resolveTodayTasks(workspaceDir: string): string[] {
     // statSync throws if the file doesn't exist; that lands in the outer catch.
     if (statSync(path).size > MAX_TASKS_MD_BYTES) return [];
     const raw = readFileSync(path, 'utf8');
-    const todayMatch = raw.match(/^##\s+(?:P\d\s*[—–-]\s*)?Today\b[\s\S]*?(?=\n##\s|$(?![\s\S]))/m);
+    const todayMatch = raw.match(/## Today[\s\S]*?(?=\n## |$)/);
     if (!todayMatch) return [];
 
     const lines = todayMatch[0].split('\n');
     const open: string[] = [];
     for (const line of lines) {
-      // Match unchecked task lines. Legacy bold form first (extracts just
-      // the task name, dropping trailing metadata), then the documented
-      // plain form (whole line body is the task).
-      const m =
-        line.match(/^\s*-\s*\[ \]\s*\*\*(.+?)\*\*/) ??
-        line.match(/^\s*-\s*\[ \]\s*(.+?)\s*$/);
+      // Match unchecked task lines: - [ ] **task name** ...
+      const m = line.match(/^\s*-\s*\[ \]\s*\*\*(.+?)\*\*/);
       if (m) open.push(sanitizeForPrompt(m[1].trim()));
     }
     return open.slice(0, 5); // cap at 5 to keep prompt lean

test/context-engine.test.ts

@@ -322,28 +322,6 @@ describe('gbrain-context engine', () => {
     expect(result.systemPromptAddition).not.toContain('Something later');
   });
 
-  it('injects documented "## P1 — Today" plain tasks from ops/tasks.md (#2186)', async () => {
-    tmpDir = makeWorkspace({
-      heartbeat: { garryAwake: true },
-      tasks: `# Tasks\n\n## P0 — Urgent\n- [ ] **Escalate outage**\n\n## P1 — Today\n- [ ] Call Alice about launch plan\n- [ ] **Review Bob contract** — due Friday\n- [x] Completed item\n\n## P2 — This Week\n- [ ] Should not surface`,
-    });
-    const engine = createGBrainContextEngine({ workspaceDir: tmpDir });
-
-    const result = await engine.assemble({
-      sessionId: 'test-session',
-      messages: [],
-    });
-
-    expect(result.systemPromptAddition).toContain('Open tasks');
-    expect(result.systemPromptAddition).toContain('Call Alice about launch plan');
-    // Bold form still extracts just the task name, not trailing metadata.
-    expect(result.systemPromptAddition).toContain('Review Bob contract');
-    expect(result.systemPromptAddition).not.toContain('due Friday');
-    expect(result.systemPromptAddition).not.toContain('Escalate outage');
-    expect(result.systemPromptAddition).not.toContain('Completed item');
-    expect(result.systemPromptAddition).not.toContain('Should not surface');
-  });
-
   it('no activity section when calendar is empty and no tasks', async () => {
     tmpDir = makeWorkspace({
       heartbeat: { garryAwake: true },

test/extract-incremental.test.ts

@@ -60,6 +60,51 @@ async function seedPage(slug: string, body: string): Promise<void> {
 }
 
 describe('runExtractCore — incremental cycle path (#417)', () => {
+  test('Dream incremental all-mode stamps the source-scoped extraction watermark (#2636)', async () => {
+    await engine.executeRaw(
+      `INSERT INTO sources (id, name, local_path) VALUES ($1, $2, $3)`,
+      ['repo-a', 'repo-a', tempDir],
+    );
+    await engine.putPage('people/alice-example', {
+      type: 'person',
+      title: 'alice-example',
+      compiled_truth: '# alice',
+      timeline: '',
+      frontmatter: {},
+      content_hash: 'h',
+    }, { sourceId: 'repo-a' });
+    writeFileSync(join(tempDir, 'people/alice-example.md'), '# alice');
+
+    await runExtractCore(engine as unknown as BrainEngine, {
+      mode: 'all',
+      dir: tempDir,
+      slugs: ['people/alice-example'],
+      sourceId: 'repo-a',
+    });
+
+    const rows = await engine.executeRaw<{ links_extracted_at: string | null }>(
+      `SELECT links_extracted_at FROM pages WHERE slug = $1 AND source_id = $2`,
+      ['people/alice-example', 'repo-a'],
+    );
+    expect(rows[0]?.links_extracted_at).not.toBeNull();
+    expect(await engine.countStalePagesForExtraction({ sourceId: 'repo-a' })).toBe(0);
+  });
+
+  test('Dream incremental dry-run does NOT stamp the watermark', async () => {
+    await seedPage('people/alice-example', '# alice');
+    await runExtractCore(engine as unknown as BrainEngine, {
+      mode: 'all',
+      dir: tempDir,
+      slugs: ['people/alice-example'],
+      dryRun: true,
+    });
+    const rows = await engine.executeRaw<{ links_extracted_at: string | null }>(
+      `SELECT links_extracted_at FROM pages WHERE slug = $1`,
+      ['people/alice-example'],
+    );
+    expect(rows[0]?.links_extracted_at ?? null).toBeNull();
+  });
+
   test('1. slugs: [] returns immediately with zero counts (early-return path)', async () => {
     await seedPage('people/alice-example', '# alice');
     const result = await runExtractCore(engine as unknown as BrainEngine, {