fix(#2474): include linters in 9c terminal condition

[fullsend-ai-coder]

PR #2468 elevated lint to a mandatory sub-step in 9c and updated five of six failure-condition sentences to say "tests or linters." The terminal condition on line 560 was missed and still read only "tests still fail," which could let an agent commit code with unresolved lint failures at retry exhaustion. Add the missing "or linters" to complete the consistency update.

---

Closes #2474

Post-script verification

• Branch is not main/master (agent/2474-fix-9c-terminal-condition)
• Secret scan passed (gitleaks — 22500115e0f3a0f55972e2c31be79a61d94942e1..HEAD)
• Pre-commit hooks passed (authoritative run on runner)
• Tests ran inside sandbox

[github-actions]

E2E tests did not run

E2E tests run automatically for org/repo members and collaborators on pull requests.

For other contributors, a maintainer must add the ok-to-test label after the latest push.

See E2E testing guide for details.

[github-actions]

Site preview

Preview: https://719ec0b6-site.fullsend-ai.workers.dev

Commit: 46eaf7d9abc763b2fe45eb95501a43dc237924e0

[fullsend-ai-review]

🤖 Finished Review · ✅ Success · Started 7:02 AM UTC · Completed 7:10 AM UTC
Commit: 5c8a7a9 · View workflow run →

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[fullsend-ai-review]

Looks good to me

---

Labels: One-word consistency fix in the scaffolded code-implementation skill (step 9c terminal condition)

Previous run

Looks good to me

---

Labels: PR fixes a bug in the scaffolded code-implementation skill (harness component)

[fullsend-ai-review]

🤖 Finished Review · ✅ Success · Started 11:58 AM UTC · Completed 12:04 PM UTC
Commit: 46eaf7d · View workflow run →

[fullsend-ai-retro]

🤖 Finished Retro · ✅ Success · Started 12:17 PM UTC · Completed 12:25 PM UTC
Commit: 46eaf7d · View workflow run →

[fullsend-ai-retro]

Retro: PR #2475 — fix(#2474): include linters in 9c terminal condition

Overview: A trivial 1-line, 2-word change to SKILL.md that added "or linters" to a terminal condition. The code agent nailed it on the first try with zero rework. However, the PR took ~5.5 hours from issue creation to merge due to infrastructure friction.

Timeline

Time (UTC)	Event
06:45	Issue #2474 created
06:46	Triage agent picks it up (1 min)
06:52	Code agent starts
06:59	PR #2475 opened (7 min code time, 1 file changed)
07:10	Review agent approves (run 1 of 6)
07:15	Human approves, adds to merge queue
07:25	❌ Merge queue failure #1 — flaky E2E (TestVendorFromSubdirectory)
08:29	Re-added to merge queue
08:44	❌ Merge queue failure #2 — same flaky E2E
11:54	Human force-pushes (identical content, new SHA)
12:01	Re-added to merge queue (attempt #3)
12:13	✅ Finally merged

What went well

• Code agent quality: Perfect first-try fix for a simple issue. No rework needed.
• Triage speed: 1 minute from issue to triage.
• Review verdict: Correct "Looks good to me" for a trivial consistency fix.

Issues identified (all already tracked)

1. Flaky E2E test caused 2 merge queue rejections (~5 hours wasted)
TestVendorFromSubdirectory fails with 422 Update is not a fast forward when concurrent merge queue entries hit shared test orgs. This is the primary reason a 1-line fix took 5.5 hours to merge.

• Already tracked: #2489

2. 6 review agent runs triggered — only 2 were legitimate
Of 6 review runs, 4 were redundant: 2 caused by self-triggering (bot's own pull_request_review submission dispatching another review), and 2 from duplicate dispatches for the same HEAD SHA.

• Already tracked: #1271 (self-triggering), #2399 (stale-head re-dispatch), #981 (concurrency groups), #1452 (same-SHA dedup)

3. No fast-path for trivial changes
A 2-word change to a markdown file consumed ~60 minutes of review agent compute across 6 runs. A lightweight review path for trivial prompt/docs changes would save significant tokens.

• Already tracked: #1900

4. Reviews still running after PR merged
Runs #5 and #6 were still in progress after the PR merged at 12:13 UTC.

• Already tracked: #2388

Conclusion

No new proposals — all identified improvements are already covered by existing open issues. The highest-impact fix would be resolving the flaky E2E test (#2489), which alone accounts for most of the wasted time. The review dedup cluster (#1271, #2399, #981) would address the token waste from redundant runs.