CI health: stale-date regression on base branch + 5 fix PRs awaiting merge (19 days open)

[erudenko]

CI Maintenance Audit — 2026-05-01

This issue was filed by the CI maintenance agent after auditing the 10 most recently failed workflow runs.

---

Summary

There are 65 cumulative failed runs in the repository. The 10 most recent failures all occurred on 2026-04-12. Every root cause has already been fixed in open PRs — but none of those PRs have been merged (19+ days open). In addition, the base branch fix/ci-type-errors-and-stale-test-date still contains stale hardcoded dates in integration.test.ts that will cause immediate test failures if CI is triggered there.

---

Root Causes (all from April 12 era)

#	Root cause	Affected files	Status
1	Stale hardcoded date "2026-03-26" in stats integration tests — outside 7/14-day rolling query windows	plugins/stats/tests/integration.test.ts	❌ Still present on base branch
2	Unused TypeScript imports (TS6133) under noUnusedLocals	tools/claudeup-core/src/services/conventions-manager.ts, doctor.test.ts, conventions-integration.test.ts	Fixed in fix PRs, not merged
3	browser-use version mismatch — plugin.json at 1.1.2, marketplace.json still at 1.1.1	.claude-plugin/marketplace.json	Fixed in fix PRs, not merged
4	Hardcoded branches allowlist in workflow triggers — any branch other than main or fix/ci-type-errors-and-stale-test-date is invisible to CI	.github/workflows/test-plugins.yml, .github/workflows/test-stats.yml	Fixed in fix PRs, not merged

---

Active Test Failures on Base Branch (fix/ci-type-errors-and-stale-test-date @ 7e440fb)

plugins/stats/tests/integration.test.ts still has "2026-03-26" as the hardcoded session date. Today (2026-05-01) that is 36 days ago, which is outside the 7-day and 14-day rolling windows used by these three tests:

Test	Query window	Result today
getSessionSummary returns correct averages and totals	7 days	0 rows → session_count assertion fails
getTopTools returns tools ranked by call count	7 days	0 rows → topTools.length > 0 fails
getDurationTrend returns daily data in ASC date order	14 days	0 rows → trend.length > 0 fails

A push to this branch or a new PR targeting it would trigger test-stats.yml and fail immediately on these three tests.

Note: db.test.ts on the base branch has already been partially fixed (getSessionSummary uses dynamic today), but integration.test.ts was not updated alongside it.

---

Open Fix PRs — Status at Time of Audit

PR	Branch	Last CI	Checks	What it fixes
#24	fix/ci-24065489776-ts-and-dates	2026-04-11	✅ 7/7 green	TS imports + dates + browser-use + partial trigger fix
#28	fix/ci-24276607217-browser-use-version	2026-04-11	GitGuardian only	browser-use version only (trigger filter blocked CI)
#31	fix/ci-main-browser-use-version	2026-04-12	GitGuardian only	targets dist-only main; test CI cannot run here
#32	fix/ci-workflow-trigger-broadening	2026-04-12	❌ 3 failures	trigger broadening only — lacked code fixes, so tests failed once CI started running
#33	fix/ci-consolidated	2026-04-12	✅ 7/7 green	All 4 root causes — TS imports + dates + browser-use + unrestricted triggers
#34	fix/ci-all-fixes-final	2026-04-12	✅ 13/13 green	integration.test.ts dates + browser-use (narrower scope than #33)

---

Recommended Action

Merge PR #33 (fix/ci-consolidated). It is the most comprehensive fix:

• All 7 CI checks green as of 2026-04-12
• Fixes all 4 root causes in a single PR
• Once merged, PRs fix(marketplace): bump browser-use version 1.1.1 → 1.1.2 in marketplace.json #28, fix(marketplace): bump browser-use version 1.1.1 → 1.1.2 to fix CI #31, fix(ci): remove hardcoded branch filters from workflow triggers #32 can be closed as superseded

After merging #33, the base branch fix/ci-type-errors-and-stale-test-date will have the rolling-date fix in integration.test.ts and the unrestricted workflow triggers, so future fix/feature branches will automatically get CI coverage without any further intervention.

---

PRs That Can Be Closed After #33 Merges

• fix(marketplace): bump browser-use version 1.1.1 → 1.1.2 in marketplace.json #28 — browser-use subset only; superseded
• fix(marketplace): bump browser-use version 1.1.1 → 1.1.2 to fix CI #31 — targets main (dist-only orphan); test CI can never run there; superseded
• fix(ci): remove hardcoded branch filters from workflow triggers #32 — trigger broadening without code fixes; failing CI; superseded by fix(ci): consolidated — TS imports, stale dates, browser-use version, unrestricted triggers #33
• fix(ci): remove unused TS imports + fix stale stats test dates #22, fix(ci): resolve merge conflict — unused TS imports + stale stats test dates #23 — earlier partial attempts; superseded by fix(ci): remove 3 unused TS imports + fix stale stats dates + re-enable CI triggers #24 / fix(ci): consolidated — TS imports, stale dates, browser-use version, unrestricted triggers #33