feat(triage): make the not-verified sentence a mechanical 2b-bis trigger

.qwen/skills/triage/references/pr-workflow.md

@@ -499,6 +499,20 @@ skipped. The value is in naming what is currently **unsubstantiated**; a
 reader who disagrees with that assessment can ignore the line, which is a
 better outcome than never seeing it.
 
+**Text trigger — this rule is mechanical, not a judgement call.** Before
+posting, grep your own draft. If your comment contains a sentence of the
+shape "not verified …", "author tested on <one platform> only", "author's
+claim, not independently re-run", or any other admission that a behavioural
+claim rests on the author's word or a single environment — **that sentence
+is the trigger.** You have already written down the gap; the 2b-bis line is
+the same sentence with the remedy attached, and omitting it means telling
+the maintainer what is missing while withholding the one command that would
+supply it. "CI is still running" does not lift the trigger: a green suite
+proves the tests pass, not that the untested behaviour holds. (Real miss
+this rule exists for: a serve-side bounded-read change whose own Stage 2
+comment said "author tested on macOS only" and never named a lane — the
+gap was written down and the remedy was not.)
+
 **Skip it — explicitly — in exactly two cases:**
 
 - **No behavioural claim to settle**: docs, types, pure refactor with an

scripts/tests/qwen-triage-workflow.test.js

@@ -553,6 +553,42 @@ describe('qwen-triage tmux workflow', () => {
     expect(order).toContain('not an enrichment');
   });
 
+  it('makes the not-verified sentence a mechanical 2b-bis trigger', () => {
+    // Post-merge measurement of #7917 (2026-07-29): 9 eligible PRs, two
+    // considered-and-declined mentions, zero positive recommendations — and
+    // the one clear behavioural candidate (#7947, bounded reads) wrote
+    // "author tested on macOS only" in its own Stage 2 comment and never
+    // named a lane. The judgement-based rule ("when 2b cannot settle it")
+    // failed exactly where the comment had already written the gap down. So
+    // the trigger is now textual: the draft's own admission is the trigger,
+    // and pending CI does not lift it.
+    const section = prSkill
+      .slice(
+        prSkill.indexOf('#### 2b-bis.'),
+        prSkill.indexOf('#### 2c. Real-Scenario'),
+      )
+      .replace(/\s+/g, ' ');
+    expect(section).toContain('mechanical, not a judgement call');
+    expect(section).toContain('grep your own draft');
+    // The trigger phrases are the ones real comments actually emit — the
+    // first two are verbatim from #7947's and #7951's Stage 2 comments.
+    expect(section).toContain('not verified');
+    expect(section).toContain('author tested on <one platform> only');
+    expect(section).toContain('not independently re-run');
+    // Pending CI must not lift the trigger: #7947's likely out was "the
+    // ubuntu Test job is still in progress".
+    expect(section).toContain('"CI is still running" does not lift');
+    // The mechanical rule must sit BEFORE the skip cases, or the skip cases
+    // read as outs from it rather than the other way around.
+    expect(section.indexOf('mechanical, not a judgement call')).toBeLessThan(
+      section.indexOf('Skip it — explicitly'),
+    );
+    // And the skip cases themselves must survive — the trigger tightens the
+    // rule, it does not replace the two legitimate outs.
+    expect(section).toContain('No behavioural claim to settle');
+    expect(section).toContain('The author lacks write access');
+  });
+
   it('names /verify on the high-risk paths, not just tmux', () => {
     // 1e is the strongest triage-time signal in the skill (10 of 31 reverted
     // PRs touched these paths vs 5 of 60 controls, p = 0.006) — and it used