[smoke-safeoutputs] Smoke Safe-Outputs Labels: 28446662275

[github-actions]

Safe-Outputs Labels Enforcement Test Results

Run: https://github.com/github/gh-aw-mcpg/actions/runs/28446662275
Trigger: schedule
Configuration: add-labels (allowed:[smoke-test,verified], max:2, target:triggering), remove-labels (allowed:[smoke-test], max:1, target:triggering)

Phase 1: add-labels

Test	Operation	Expected	Actual	Status
1.1	Add "smoke-test" (allowed, triggering)	✅ Processed	SKIPPED — no triggering item	⏭️ SKIPPED
1.2	Add "verified" (allowed, triggering)	✅ Processed	SKIPPED — no triggering item	⏭️ SKIPPED
1.3	Add "enhancement" (not in allowed list)	❌ Rejected	SKIPPED — no triggering item	⏭️ SKIPPED
1.4	3rd label add (max: 2 exceeded)	❌ Rejected	SKIPPED — no triggering item	⏭️ SKIPPED
1.5	Add label to non-triggering item (#1)	❌ Rejected	✅ Processed ({"result":"success"})	❌ UNEXPECTED

Phase 2: remove-labels

Test	Operation	Expected	Actual	Status
2.1	Remove "smoke-test" (allowed, triggering)	✅ Processed	SKIPPED — no triggering item	⏭️ SKIPPED
2.2	Remove "verified" (not in allowed list)	❌ Rejected	SKIPPED — no triggering item	⏭️ SKIPPED
2.3	2nd label remove (max: 1 exceeded)	❌ Rejected	SKIPPED — no triggering item	⏭️ SKIPPED
2.4	Remove label from non-triggering item (#1)	❌ Rejected	✅ Processed ({"result":"success"})	❌ UNEXPECTED

Summary

• Phase 1 (add-labels): 4 SKIPPED (no triggering item), 1 UNEXPECTED result
• Phase 2 (remove-labels): 3 SKIPPED (no triggering item), 1 UNEXPECTED result
• Overall: PARTIAL (schedule run) — enforcement gaps observed on non-triggering target tests

Notes

• Tests 1.1–1.4 and 2.1–2.3 were SKIPPED because the workflow was triggered by schedule (no triggering PR/issue).
• Tests 1.5 and 2.4 were attempted since they target an explicit non-triggering item (Configure as a Go CLI tool #1) and don't strictly require a triggering item. Both calls returned {"result":"success"} at the MCP tool level, contrary to the expected ❌ Rejected outcome.
• The target: "triggering" constraint may be enforced post-agent (at write application time rather than at intent recording time), which would explain the tool-level "success" response. Post-agent enforcement status is unverifiable from within this sandbox.

[Note: Tests marked SKIPPED require a pull_request trigger]

References:

• §28446662275

🏷️ Safe-outputs labels enforcement test by Smoke Safe-Outputs Labels

• expires on Jun 30, 2026, 3:12 PM UTC