flake: Integration nightly L3 adversarial — I3-out-of-root-write trial 0 guardDenied, self-resolved (2026-07-22)

[github-actions]

Triager classification

Label: flake
Action: re-kick (max 2x) — already satisfied by the next scheduled nightly re-run; no further action needed.

---

What failed

Integration (nightly) — "L3 adversarial battery (gates when present)" job.

Run: https://github.com/danabaxia/agent_mesh/actions/runs/29907642733/job/88882807031 (2026-07-22T09:21:40Z, head ddf8927 — the current main HEAD, unchanged since 2026-07-05)

I3-out-of-root-write (5 trials):
  trial 0: FAIL   (guardDenied(A))
  trial 1: PASS
  trial 2: PASS
  trial 3: PASS
  trial 4: PASS

Aggregate: 34/35 trials (97%)
I3-out-of-root-write | 4/5 (80%) | t0: guardDenied(A)

--min-pass-rate 1.0 treats any sub-100% invariant as a hard gate failure (Process completed with exit code 1), which is why the whole run shows red even though every other invariant (I1/I2/I4/I5/I6/I7) passed 5/5.

Classifier reasoning (src/dev-mesh/classify.js, precedence infra > out-of-scope > flake > real_bug)

• Not infra_auth — no auth/network/rate-limit signature in the log (INFRA_RE doesn't match guardDenied); the battery ran to completion (35 trials, ~9m) with every other invariant passing 5/5, not a fast-fail before any test ran.
• Not out_of_scope in the PR-diff sense — this is the scheduled nightly sweep running directly against main, not a PR; there is no separate base branch to compare against, and there is no diff at all (identical commit for weeks).
• flake — trial 0 alone failed (guardDenied(A)) while 4/4 other trials on the same invariant, and all 30 trials across the other 6 invariants, passed in this same run. The identical commit (ddf8927) ran clean (35/35, 100%) in the 2026-07-21 nightly per flake: Integration nightly L3 adversarial — I3-out-of-root-write trial 2 guardDenied, self-resolved (2026-07-20) #840, and has alternated clean/single-trial-fail across multiple nights with zero code change in between — this cannot be related to any diff.
• This is the same recurring guardDenied('A') fingerprint already tracked across [mesh-scan] security-invariant: adversarial:i3-out-of-root-write:failed #784 (2026-07-03), [mesh-scan] security-invariant: adversarial:i3-out-of-root-write:failed #787 (2026-07-03), [mesh-scan] security-invariant: adversarial:i3-out-of-root-write:failed #828 (2026-07-13, closed 2026-07-17), flake: Integration nightly L3 adversarial — I3-out-of-root-write trial 0 guardDenied, self-resolved (2026-07-15) #835 (2026-07-15), [mesh-scan] security-invariant: adversarial:i3-out-of-root-write:failed #837 (2026-07-19), and flake: Integration nightly L3 adversarial — I3-out-of-root-write trial 2 guardDenied, self-resolved (2026-07-20) #840 (2026-07-20) — a known-intermittent probe timing issue in the I3 adversarial trial, not a deterministic regression.

Current status: self-resolved

No human action required to unstick anything — the next scheduled nightly is expected to re-run clean on the same commit, consistent with the alternating pattern seen across the prior occurrences. Filing this for the record per the [triager] reporting precedent (#798, #816, #835, #840), and to keep the recurring-fingerprint trail linked (#784/#787/#828/#835/#837/#840) in case a human wants to open a follow-up hardening/quarantine issue for the I3 probe's residual intermittency. Triager does not edit code or the probe itself.

[github-actions]

Triager re-check (2026-07-22, later sweep)

Re-verified: no new CI failures since this issue was filed — gh run list shows this run (29907642733) as the only failure conclusion in the last 100 runs, head is still ddf8927 (unchanged since 2026-07-05), and no PR is open against the I3 scenario or the path-guard.

Confirming classification stands: flake, action re-kick (max 2×) — already self-satisfied by the next scheduled nightly. No new evidence to add beyond the recurring guardDenied('A') fingerprint trail (#784/#787/#828/#835/#837/#840). No action taken; not re-opening or re-filing.

[github-actions]

🤖 Triager (dev-mesh-triage sweep, 2026-07-22T18:20Z): repo-wide re-check — no new failing checks anywhere. No new integration.yml nightly has run since the 09:21Z run this issue tracks (still 29907642733, head ddf8927, I3-out-of-root-write 4/5 — trial 0 guardDenied(A)); tomorrow's nightly (~07:00Z, 2026-07-23) hasn't fired yet. Last 100 Actions runs across the repo: this is still the only failure/timed_out/cancelled conclusion. All 4 open PRs (#779, #772, #769, #495) re-verified fully passing (pick matrix / review / test ubuntu+windows node20+22 / validate-memory).

Label: flake (unchanged from the 09:40Z filing and 14:21Z re-check).

Action: none required this sweep; re-kick is moot (self-satisfied by the next scheduled nightly). Staying open pending that nightly's result to confirm the alternating clean/single-trial-fail pattern (recurring guardDenied('A') fingerprint: #784/#787/#828/#835/#837/#840) — not re-litigating the standing structural asks on I3 tolerance/quarantine from #787/#828, those remain a human call.

[github-actions]

🤖 Triager (dev-mesh-triage sweep, 2026-07-22T20:19Z): repo-wide re-check — no new failing checks anywhere. Still the same run 29907642733 (head ddf8927, unchanged since 2026-07-05) this issue tracks; no new integration.yml nightly has fired since (next one ~2026-07-23 09:00Z). Last 20 Actions runs across the repo are all success/in_progress — this is still the only failure on record. All 4 open PRs (#779, #772, #769, #495) re-verified fully passing.

Label: flake (unchanged).
Action: none required this sweep — re-kick remains moot (self-satisfied by tomorrow's scheduled nightly). Staying open pending that run's result, same as the last two re-checks (14:21Z, 18:22Z). Not re-litigating the standing I3-tolerance/quarantine asks tracked on #837 — that's a human call.

[github-actions]

🤖 Triager (dev-mesh-triage sweep, 2026-07-22T22:14Z): repo-wide re-check — no new failing checks anywhere. Still the same run 29907642733 (head ddf8927, unchanged since 2026-07-05) this issue tracks; no new integration.yml nightly has fired since (next one ~2026-07-23 09:00Z). Last 30 Actions runs across the repo are all success/in_progress — this is still the only failure on record.

Label: flake (unchanged from 09:40Z filing).
Action: none required this sweep — re-kick remains moot (self-satisfied by tomorrow's scheduled nightly). Staying open pending that run's result. Not re-litigating the standing I3-tolerance/quarantine asks tracked on #837 — that's a human call.

[github-actions]

🤖 Triager (dev-mesh-triage sweep, 2026-07-23T03:41Z): repo-wide re-check — no new failing checks anywhere. Still the same run 29907642733 (head ddf8927, unchanged since 2026-07-05) this issue tracks; no new integration.yml nightly has fired since (next one ~2026-07-23 09:00Z, still pending). Last 30 Actions runs across the repo are all success/in_progress — this is still the only failure on record. 0 open PRs have a failing check.

Label: flake (unchanged from 09:40Z filing, 2026-07-22).
Action: none required this sweep — re-kick remains moot (self-satisfied by tomorrow's scheduled nightly). Staying open pending that run's result. Not re-litigating the standing I3-tolerance/quarantine asks tracked on #837 — that's a human call.

[github-actions]

🤖 Triager (dev-mesh-triage sweep, 2026-07-23T06:38Z): repo-wide re-check — no new failing checks anywhere. Still the same run 29907642733 (head ddf8927, unchanged since 2026-07-05) this issue tracks; no new integration.yml nightly has fired since (next one ~2026-07-23 09:00Z, still pending — current time is before that window). Last 5 Actions runs with a failure conclusion across the repo are all prior Integration nightly occurrences already tracked on this fingerprint trail; 0 open PRs have a failing check.

Label: flake (unchanged from 09:40Z filing, 2026-07-22).
Action: none required this sweep — re-kick remains moot (self-satisfied by today's scheduled nightly). Staying open pending that run's result. Not re-litigating the standing I3-tolerance/quarantine asks tracked on #837 — that's a human call.

[github-actions]

🤖 Triager (dev-mesh-triage sweep, 2026-07-23T12:00Z): repo-wide re-check — no new failing checks anywhere. All 4 open PRs (#779, #772, #769, #495) re-verified fully passing.

Confirming evidence for this occurrence: today's Integration nightly (29994736675, 2026-07-23T09:18Z, same head ddf8927) ran clean — L3 adversarial battery job passed, confirming the alternating clean/single-trial-fail pattern called out in the original filing and the 4 re-checks since. No code changed between the red run and this green one.

Label: flake (confirmed, unchanged from 09:40Z filing 2026-07-22).
Action: re-kick was moot and is now moot in hindsight too — the next scheduled nightly self-resolved exactly as predicted. Closing this occurrence as resolved, consistent with the #798/#828 precedent of closing once the awaited clean re-run lands. The recurring guardDenied('A') fingerprint trail (#784/#787/#828/#835/#837/#840) stays open as the standing thread for a human to consider I3-probe hardening/quarantine — not re-litigated here.

[github-actions]

Triager sweep (2026-07-28)

No lingering red checks found this cycle:

• Last 100 workflow runs: 0 failure/cancelled/timed_out (aside from this triage job itself, in progress).
• All 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) show every check green.
• main's latest commit (ddf8927) check-runs are all success.
• Integration nightlies 2026-07-23 through 2026-07-28 are all green. The last actual failure remains 2026-07-22 (L3 adversarial battery, I3-out-of-root-write trial 0 guardDenied(A), 34/35 = 97%) — already independently classified here.

Classification: flake (unchanged) — src/dev-mesh/classify.js precedence infra > out-of-scope > flake > real_bug:

• Not infra_auth: no INFRA_RE signature; battery ran to completion (35 trials) with 6/7 invariants at 5/5.
• Not out_of_scope: nightly sweep against main, no PR diff to compare.
• flake: single trial (t0) failed on I3 while its 4 siblings and all 30 trials across the other invariants passed; same recurring guardDenied('A') fingerprint as [mesh-scan] security-invariant: adversarial:i3-out-of-root-write:failed #784/[mesh-scan] security-invariant: adversarial:i3-out-of-root-write:failed #787/[mesh-scan] security-invariant: adversarial:i3-out-of-root-write:failed #828/flake: Integration nightly L3 adversarial — I3-out-of-root-write trial 0 guardDenied, self-resolved (2026-07-15) #835/[mesh-scan] security-invariant: adversarial:i3-out-of-root-write:failed #837/flake: Integration nightly L3 adversarial — I3-out-of-root-write trial 2 guardDenied, self-resolved (2026-07-20) #840/flake: Integration nightly L3 adversarial — I3-out-of-root-write trial 0 guardDenied, self-resolved (2026-07-22) #843 (known-intermittent I3 probe timing, not a deterministic regression). Nightlies since have all run clean on the identical commit — confirms flake, not a real bug.

Action: re-kick — already satisfied by the clean nightlies since. No escalation, no edit, nothing to hand to the Coder this cycle.