[aw-failures] P2: Daily-AIC guardrail marks budget-blocked runs as `failure` (core.setFailed) — false-failure noise + auto-issue [Content truncated due to length]

[github-actions]

Recommendation

Stop the daily-AIC guardrail from calling core.setFailed when the budget is exceeded — emit a neutral/skipped conclusion instead. The agent job is already gated on the daily_ai_credits_exceeded output, so the setFailed is redundant for control flow and only serves to paint budget-blocked runs red — which dominates the failure signal and spawns per-workflow auto-failure issues.

Problem statement

In the 2026-06-13 13:17Z 6h window, 14 of 20 "failed" agentic runs were not real failures — they were scheduled reviewer workflows correctly blocked by the daily-AIC guardrail after the shared 5000 daily-AIC budget was exhausted (~11:00Z). Each blocked run exits the activation job via core.setFailed, so the run conclusion is failure even though the design intent is "skip this run." Side effects: (1) reliability/failure dashboards over-count failures, (2) auto-failure issues are filed per workflow (#39059 Test Quality Sentinel, #39050 Matt Pocock Skills Reviewer), and (3) this Failure Investigator is triggered on non-actionable noise.

Root cause

The guardrail script hard-fails the step on exceedance:

• actions/setup/js/check_daily_aic_workflow_guardrail.cjs:593 → core.setFailed(\Daily workflow AIC guardrail exceeded for ${workflowName}: ${totalAIC}/${threshold}.`)`
• It also already sets the control output at line 586 → core.setOutput("daily_ai_credits_exceeded", "true").

The downstream agent job is gated on that output, independent of step status:

• pkg/workflow/compiler_main_job.go:69 → needs.activation.outputs.daily_ai_credits_exceeded != 'true'

Observed guardrail log (representative run §27465717773):

##[warning]Daily workflow AIC guardrail exceeded for Test Quality Sentinel: 5017.38/5000.
##[error]Daily workflow AIC guardrail exceeded for Test Quality Sentinel: 5017.38/5000.

Because the agent job already keys on daily_ai_credits_exceeded, the core.setFailed at line 593 is redundant for skipping the agent — its only effect is the red failure conclusion.

Affected workflows and run IDs

Workflow	Runs (in-window)
Test Quality Sentinel	§27465717773, §27465421432, §27465299831, §27464789069, §27464680872
PR Code Quality Reviewer	§27465717777, §27465421426, §27465299828, §27464789060, §27464680882
Matt Pocock Skills Reviewer	§27465421436, §27464789067, §27464680862, §27463928154

Representative run: §27465717773 (clearest 5017.38/5000 exceedance log). Comparator: same workflows succeeded earlier in the day before the shared budget was exhausted; no in-window successful comparator exists because every post-exhaustion run is blocked.

Why this is distinct from existing AIC issues

• [perf-improvement] Raise max-ai-credits for Failure Investigator (6h) — meta-monitor blind spot #38624 (raise max-ai-credits for Failure Investigator) and [deep-report] Add a soft pre-cap AI-credits guard to heavy aggregator workflows #38645 (graceful aggregator degradation at the AIC cap) address budget consumption/tuning.
• This issue is about how a budget-blocked run is reported — it should conclude skipped/neutral, not failure — independent of what the budget is set to. Fixing the budget does not remove the false-failure noise.

Proposed remediation

1. In check_daily_aic_workflow_guardrail.cjs, on exceedance keep core.setOutput("daily_ai_credits_exceeded", "true") and core.warning(...), but replace core.setFailed(...) (line 593) with a non-failing exit so the activation job concludes success (the agent still skips via the output gate).
2. Optionally surface the block as a neutral GitHub conclusion (e.g. a notice + a step-summary line) so the run is visibly "budget-skipped," not green-as-if-ran.
3. Suppress the per-workflow auto-failure issue path ([aw] Test Quality Sentinel exceeded daily AI credits budget #39059/[aw] Matt Pocock Skills Reviewer exceeded daily AI credits budget #39050 class) when daily_ai_credits_exceeded == 'true', since the run is a deliberate skip, not a failure.

Success criteria / verification

1. A workflow blocked by the daily-AIC guardrail produces a run whose conclusion is not failure (skipped/neutral/success-with-skip).
2. The agent job still does not execute when daily_ai_credits_exceeded == 'true' (existing gate unchanged; verify via daily_aic_workflow_guardrail_test.go).
3. No [aw] ... exceeded daily AI credits budget auto-failure issue is filed for a guardrail-skipped run.
4. A subsequent Failure Investigator window over a budget-exhaustion period reports these runs as skipped, not as P-tier failures.

References

• Representative: §27465717773 · this analysis run §27467779429
• Spawned auto-issues: [aw] Test Quality Sentinel exceeded daily AI credits budget #39059, [aw] Matt Pocock Skills Reviewer exceeded daily AI credits budget #39050 · Related budget issues: [perf-improvement] Raise max-ai-credits for Failure Investigator (6h) — meta-monitor blind spot #38624, [deep-report] Add a soft pre-cap AI-credits guard to heavy aggregator workflows #38645, [agentic-token-audit] Daily AIC Usage Audit — 2026-06-12 #38834
  Related to Workflow Health Manager - Meta-Orchestrator - Issue Group #29109

Generated by 🔍 [aw] Failure Investigator (6h) · 191.6 AIC · ⌖ 12.9 AIC · ⊞ 5.1K · ◷

• expires on Jun 20, 2026, 5:26 AM UTC-08:00

[github-actions]

🍪 Issue Monster selected this for Copilot

I've identified this issue as a good candidate for automated resolution and requested assignment to the Copilot coding agent.

If assignment succeeds, the Copilot coding agent will analyze the issue and create a pull request with the fix.

Om nom nom! 🍪

🍪 Om nom nom by Issue Monster · 17.8 AIC · ⌖ 4.37 AIC · ⊞ 22.3K · ◷

[copilot-swe-agent]

The agent encountered an error and was unable to start working on this issue: Please try again later, or contact support if the problem persists. (Request id: 1A40:3FC6F0:22A9CE6:80EA5A6:6A2D5EED)