needs-human: PR#772 stuck (not-clean:UNSTABLE) — non-required check unmet

[danabaxia]

🤖 dev-mesh ② backstop: PR#772 is UNSTABLE — GitHub considers it technically mergeable, but a non-required (or pending/skipped/unreported) status check is unmet. The content fixers have no handle on this: there is no merge conflict for dev-mesh-mergefix to resolve and no failing required check for dev-mesh-ci-sweep to repair. A human needs to resolve the CI/branch-protection state — fix or re-run the offending check, or drop it from the required set — then merge.

• item: automerge:PR#772
• detail: not-clean:UNSTABLE

[danabaxia]

🔬 Analyst research (ask):

I need permission to use WebSearch to research this failure pattern — can you grant it so I can continue the diagnosis?

[danabaxia]

🛠 Draft fix (do-mode, never auto-merged): PR #779. Review, un-draft, and merge if good.

[github-actions]

Triager report

Label: infra_auth

Classifier reason (src/dev-mesh/classify.js precedence: infra > out-of-scope > flake > real_bug): the CI workflow run for PR #772's head commit (9e866b2) completed with conclusion: action_required and zero jobs executed — gh api .../check-runs returns no check-runs at all for that SHA. That's the !ranAnyTest branch ('no test executed before exit'), which the classifier ranks above flake/real_bug regardless of any diff content. Not out_of_scope (nothing reproduces — nothing ran). Not flake (there's no run to re-kick into passing; action_required isn't a test failure, it's Actions withholding the run).

Root cause (why it's stuck): PR #772 was opened by github-actions[bot] (the framework's own automation), and its CI pull_request run is sitting in action_required with 0 jobs — GitHub is withholding automatic execution for a run whose trigger identity isn't a trusted human/collaborator context. This is the same class of problem as #771 (missing DEV_MESH_PAT, bot-token operations hitting platform-side trust gates) — a repo/Actions configuration gap, not application code.

Action: escalate to a human (already correctly done — this issue exists for exactly that). Confirming the existing escalation is correctly classified and does not need a code fix on main. Note PR #779 (draft, do-not-merge, already reviewed/approved) is a different, valid fix — it stops the research-escalation loop from wasting dispatches trying to diagnose/fix this specific unfixable-by-code case — but it does not and cannot unstick #772 itself. Unsticking #772 requires a human to either manually approve/re-run the withheld workflow run or adjust the branch-protection/required-check configuration so bot-authored PRs aren't gated this way.

[github-actions]

Triager sweep (hourly, 2026-07-31): re-checked the CI state referenced here.

• PR memory:promote — production-vs-fixture-precision-fix #772 mergeStateStatus is now CLEAN (was UNSTABLE); every check (pick matrix, review, test ×4, validate-memory) is SUCCESS.
• Swept all 4 open PRs (research-fix: needs-human: PR#772 stuck (not-clean:UNSTABLE) — non-required check unmet (#778) #779, memory:promote — production-vs-fixture-precision-fix #772, memory:promote — sibling-metric-noise-band-audit (PR #761 resolution) #769, memory(curator): expire regression-band-override-semantics (valid_to 2026-06-23) #495) — all CLEAN, no failing/pending/unreported checks on any.
• Swept recent workflow runs (scheduled dev-mesh pipelines + integration.yml nightly) — all green; the last integration.yml failure was 2026-07-22 and is stale/already handled.

No lingering red check to classify right now (classifier: src/dev-mesh/classify.js, precedence infra > out-of-scope > flake > real_bug — n/a, nothing to run it against). Recommending this issue be closed as resolved; re-open if the UNSTABLE state recurs.

[github-actions]

Triager sweep (hourly, 2026-08-01): re-checked CI state.

• Swept last 100 workflow runs: conclusions are exclusively success (or in-progress for this very sweep) — no failure/cancelled/timed_out runs.
• Swept all 4 open PRs (research-fix: needs-human: PR#772 stuck (not-clean:UNSTABLE) — non-required check unmet (#778) #779, memory:promote — production-vs-fixture-precision-fix #772, memory:promote — sibling-metric-noise-band-audit (PR #761 resolution) #769, memory(curator): expire regression-band-override-semantics (valid_to 2026-06-23) #495) — all mergeable/CLEAN, every check (pick matrix, review, test ×4, validate-memory) is SUCCESS.
• Checked check-runs on main's HEAD (ddf8927) — all completed/success except this sweep's own in-progress triage run.
• No new CI-failure issues filed since the last sweep (2026-07-31).

No lingering red check to classify right now (classifier: src/dev-mesh/classify.js, precedence infra > out-of-scope > flake > real_bug — n/a, nothing to run it against). Reiterating the prior recommendation: this issue can be closed as resolved; re-open if the UNSTABLE state recurs on #772 or elsewhere.