the-own-lab/verify

verify

Purpose

Close the loop between spec and code. Run the test suite, then match TODO/REQ acceptance criteria to actual passing tests. verify ≠ tdd: tdd writes one test at a time; verify inspects the whole delivered state before update-docs.

Inputs

• BRIEF.md ## Current Contract, ### Requirements Summary, and ### TODO In Scope (requirement list, acceptance criteria, and scoped TODOs already folded in by read-brief).
• BRIEF.md ## Test Plan (produced earlier by test-plan and tdd).
• The actual test command(s) for the project (derive from package.json, project.json, mise.toml, or ask user if unclear).
• Does NOT re-open REQUIREMENTS.md / TODO.md itself; the brief is authoritative.

Outputs

• Pass/fail summary written to the brief under ## Assumptions as verify: all REQs mapped to passing tests or a list of gaps.
• If gaps found, items appended to ## Open Questions for follow-up.

Procedure

1. Run unit tests via pnpm nx test <project> (or equivalent). Capture result.
2. Run e2e/integration tests if the project defines them.
3. For every REQ summarized in the brief's ### Requirements Summary, locate at least one test exercising its acceptance.
4. For every TODO item the brief records as in-flight for this run, confirm the code actually delivers it.
5. Apply the security lens (absorbed from the retired security-scan skill):
   • new user-input boundaries without validation?
   • new secrets / env vars / tokens in committed code?
   • new file-path / shell-exec / SQL construction from user data? Flag findings as plausible risks, not confirmed vulns, unless the evidence is explicit.
6. If any check fails, leave the work incomplete; do NOT proceed to update-docs.

Failure modes

• Test command fails due to missing runtime → stop and surface the mise/pnpm issue rather than marking verify pass.
• REQ has no test coverage → gap; append to ## Open Questions. User decides whether to add a test now or defer.
• Security lens produces > 3 findings → likely systemic; recommend a full /review.

Does not do

• Does not write new tests (that is tdd).
• Does not decide WHAT to test (that is test-plan).
• Does not tick TODO boxes (that is update-docs).
• Does not replace /review Layer 2 red-team analysis.