Suppress noisy auto-filed failure issues for recurring AI-credits failures

[Evangelink]

Closes #9109
Closes #9102
Closes #9093
Closes #9086
Closes #9085
Closes #9074

These agentic workflows were auto-filing noisy failure issues when they exhausted AI credits after already producing their useful output. This PR disables failure issue reporting with safe-outputs.report-failure-as-issue: false for the recurring perf, efficiency, test, review, and grade-tests workflows, matching the precedent already used by build-failure-analysis*.md.

Bumping max-ai-credits was considered but rejected because the workflows still produce useful PRs (for example #9108, #9101, and #9092) before running out. Only the cleanup step crashes, and that step does not need to be retried via auto-filed tracking issues.

[Copilot]

Pull request overview

Disables auto-filing GitHub issues for agentic workflow failures that occur after useful output is already produced (notably recurring AI-credits exhaustion), by setting safe-outputs.report-failure-as-issue: false across the affected workflows in this repo.

Changes:

• Added safe-outputs.report-failure-as-issue: false to the workflow source files for perf/efficiency/test improvers, PR review, and grade-tests.
• Regenerated the corresponding *.lock.yml files so the compiled workflows set GH_AW_FAILURE_REPORT_AS_ISSUE: "false".

Show a summary per file

File	Description
.github/workflows/test-improver.md	Disables auto-filing failure issues via safe-outputs.report-failure-as-issue: false.
.github/workflows/test-improver.lock.yml	Compiled output updated to set GH_AW_FAILURE_REPORT_AS_ISSUE: "false".
.github/workflows/review-on-open.agent.md	Disables auto-filing failure issues via safe-outputs.report-failure-as-issue: false.
.github/workflows/review-on-open.agent.lock.yml	Compiled output updated to set GH_AW_FAILURE_REPORT_AS_ISSUE: "false".
.github/workflows/perf-improver.md	Disables auto-filing failure issues via safe-outputs.report-failure-as-issue: false.
.github/workflows/perf-improver.lock.yml	Compiled output updated to set GH_AW_FAILURE_REPORT_AS_ISSUE: "false".
.github/workflows/grade-tests-on-pr.agent.md	Disables auto-filing failure issues via safe-outputs.report-failure-as-issue: false.
.github/workflows/grade-tests-on-pr.agent.lock.yml	Compiled output updated to set GH_AW_FAILURE_REPORT_AS_ISSUE: "false".
.github/workflows/efficiency-improver.md	Disables auto-filing failure issues via safe-outputs.report-failure-as-issue: false.
.github/workflows/efficiency-improver.lock.yml	Compiled output updated to set GH_AW_FAILURE_REPORT_AS_ISSUE: "false".

Copilot's findings

• Files reviewed: 10/10 changed files
• Comments generated: 0

[Evangelink]

Note

🤖 Automated review by GitHub Copilot. Posted via a maintainer's GitHub token, so it appears under their account — the account owner did not write or approve this content personally. Generated by the Expert Code Review workflow. To request a follow-up action, reply by tagging @copilot directly.

Review Summary

This is a clean, well-scoped configuration change. All 5 affected workflows are updated consistently: both the .md source file and the corresponding compiled .lock.yml file are modified in every case, as required by the agentic workflow guidelines.

Lock File Integrity Verification

All lock file metadata hashes are correct:

• frontmatter_hash is updated for every workflow (expected — frontmatter changed).
• body_hash is unchanged for every workflow (correct — only safe-outputs frontmatter changed, not the agent body/prompt).
• "strict":true is preserved in all five lock files. ✅

Minor Observation: grade-tests.agent.md not in scope

grade-tests.agent.md (the /grade-tests slash-command variant) still carries GH_AW_FAILURE_REPORT_AS_ISSUE: "true" in its lock file and has no report-failure-as-issue: false in its source. It already has noop.report-as-issue: false, which is a partial mitigation.

Since that workflow is manually triggered (not recurring/scheduled), omitting it here is likely intentional — a human issued the command so the failure is more expected and visible. If the same AI-credit exhaustion pattern is ever observed from a manual /grade-tests run, it would be easy to follow up with a single-line addition there too.

---

22-Dimension Verdict Table

#	Dimension	Status	Notes
1	Correctness	✅ Pass	report-failure-as-issue: false in .md compiles to GH_AW_FAILURE_REPORT_AS_ISSUE: "false" in lock; all five pairs aligned
2	Security	✅ N/A	No security surface changed
3	Performance	✅ N/A	Configuration only
4	Error handling	✅ Pass	Failure mode intentionally changed from auto-issue to silent; deliberate and well-reasoned in the PR description
5	Edge cases	✅ N/A	No code logic involved
6	Code style	✅ Pass	YAML formatting is consistent with existing safe-outputs: blocks
7	Testing	✅ N/A	No tests applicable for workflow config
8	Documentation	✅ N/A	No documentation update required
9	Breaking changes	✅ Pass	Reduces noise; no consumer-visible contract broken
10	Public API surface	✅ N/A	Not applicable
11	Thread safety	✅ N/A	Not applicable
12	Memory / resource leaks	✅ N/A	Not applicable
13	Logic bugs	✅ Pass	No logic introduced
14	Naming conventions	✅ Pass	Field name matches the gh aw schema (report-failure-as-issue)
15	Dependencies	✅ N/A	No dependency changes
16	Scope discipline	✅ Pass	Only report-failure-as-issue touched; GH_AW_MISSING_TOOL_REPORT_AS_FAILURE and GH_AW_MISSING_DATA_REPORT_AS_FAILURE deliberately left as "true"
17	Consistency	✅ Pass	All five workflows updated uniformly
18	Intent alignment	✅ Pass	Directly suppresses the auto-filed issues referenced in the PR (closes #9085, #9086, #9093, #9102, #9109, #9074)
19	Lock file integrity	✅ Pass	frontmatter_hash updated; body_hash unchanged; compiler_version v0.79.4 and engine_versions preserved
20	Compilation workflow compliance	✅ Pass	Each .md source change is accompanied by its compiled .lock.yml; strict mode maintained
21	Strict mode preserved	✅ Pass	"strict":true in all five lock file metadata headers
22	Coverage completeness	⚠️ Minor	grade-tests.agent.md (slash-command variant) is not in scope; likely intentional (manual trigger ≠ recurring), but worth keeping in mind if that workflow also starts filing noisy issues

Overall verdict: ✅ LGTM — no blocking or major issues found. The change is correct, minimal, well-motivated, and matches the established precedent from build-failure-analysis*.md.

🤖 Automated content by GitHub Copilot. Posted via a maintainer's GitHub token, so it appears under their account — the account owner did not write or approve this content personally. Generated by the Expert Code Review (on PR ready) workflow. · 353.3 AIC · ⌖ 12.3 AIC · ◷