[codex] Guardrail backlog: Test coverage gap (test-coverage-gap)

[haasonsaas]

Test coverage gap

This issue tracks a recurring review-feedback class from the EvalOps review feedback sentinel.

• Class: test-coverage-gap
• Score: 80
• Findings: 1
• Repos: evalops/maestro-internal
• Generated at: 2026-05-23T07:15:53Z
• Window: merged since 2026-05-20 with minimum severity high

Guardrail to build

Add focused regression or fuzz coverage for the exact missed case, then wire it into the smallest relevant CI target.

Representative feedback

• p1 evalops/maestro-internal#2103 test/scripts/ci-guardrails.test.ts:405
  • Keep internal runner guardrail tied to repo identity
  • https://github.com/evalops/maestro-internal/pull/2103#discussion_r3291910563

Finding fingerprints

• 5f154356d2776b24c8d90c2dc275da5c40366b3cd2c45392edcee956fc69a885

Acceptance criteria

• The class has an owner repo and a concrete guardrail location.
• The guardrail fails for at least one representative feedback shape listed above.
• The guardrail is wired into the smallest relevant CI or preflight target.
• The issue is closed only after the guardrail has merged and the feedback sentinel no longer ranks this class as an unaddressed candidate.

[haasonsaas]

Closed after the representative Agent PM coverage gap was fixed and verified.

Evidence:

• test: include pyproject in Bazel pytest data agent-pm#21 merged the Bazel test data guardrail by including pyproject.toml in SERVICE_DATA.
• build: refresh uv lock annotations for Bazel check agent-pm#22 merged the follow-up lock/remote-execution fix after CI went green: Bazel RBE smoke, CI lint, typecheck, storage-tests, and test all completed successfully.
• Final org review-thread audit returned [] for open PRs.

[haasonsaas]

Closing because the review feedback sentinel no longer ranks test-coverage-gap as an active guardrail candidate in the current backlog window.