| name | verify-checklist |
| description | Verify phase — confirm the change actually works: tests run, gate passes, acceptance criteria met, claims grounded in output. |
verify-checklist
You are in the verify phase of Coding Mode. Confirm the change does what it
claims — by evidence, not assertion.
Walk the checklist and report the result of each:
- Tests. Did you run them, and what was the output? A passing run is the
evidence; "should pass" is not.
- Quality gate. Did
./scripts/agent-quality-gate.sh (fmt, test, vet, lint)
pass? Report failures verbatim.
- Acceptance criteria. Walk each criterion in the ticket and point at the
test or behaviour that satisfies it.
- No regressions. Did the side effects found in analyse get covered?
Grounding (required)
Every "done" must point at evidence — a test name, a command's output, a
file/line. Claiming success without a concrete anchor is a verify failure.
Never report "looks good" or "should be fine".
Good (grounded):
go test ./internal/codingskills/ passes (4 tests); the Pack() count test
pins the five-skill set.- Acceptance "zero cost when off" met:
TestPhaseSkillsOnlyWhenActive asserts
no skill block is appended without an active mode.
Bad (rejected — no anchor):
- Everything works.
- Tests should be passing now.
