verify

Name: verify
Author: eduwxyz

15stars1forks

How to install verify?

eduwxyz/my-awesome-skills/verify

$npx -y skills add eduwxyz/my-awesome-skills --skill verify

Installs into the current project.

Summary

Verify that a spec.md's Acceptance criteria are observably covered by green tests before opening a PR. Reads the spec, finds the tests that cover each AC, runs the suite to confirm everything green, and reports gaps (uncovered AC, skipped tests, missing edge-case coverage). When

Files · 1

View on GitHub

SKILL.md

1	# Verify
2
3	The closing gate of the SDD pipeline. Map every acceptance criterion in the spec to a green test, surface any gaps, decide whether the work is ready for review.
4
5	## Skip verify for
6
7	- Spike / scratch work — there's no spec to verify against.
8	- Tiny obvious changes — typo fix, single-line bug, comment update.
9	- Specs without an `## Acceptance criteria` section — the upstream skill (`interview-to-spec` or `diagnose`) was abandoned mid-way. Finish that first.
10
11	## Before you start
12
13	Identify the spec. A path or slug for an existing `spec/<slug>.md`. If neither was provided, ask the user — do not guess.
14
15	Validate it. The spec must contain `## Acceptance criteria`. If missing, abort and tell the user to complete the spec first.
16
17	## The work
18
19	1. Read the spec. Focus on `Acceptance criteria` and `Edge cases` (or `Related risks` for bug specs).
20
21	2. Find the tests. Locate the relevant test files in the project's test directory (`tests/`, `__tests__/`, `_test.go`, `_spec.rb`, etc). Read them.
22
23	3. Map AC → test. For each acceptance criterion, find the test that observably proves it. Match on behaviour, not on naming similarity. A test named close to an AC does not count if it asserts something different.
24
25	4. Find gaps. Two categories — handled differently:
26
27	Blocking gaps (auto-fixed via handoff to `tdd`):
28	- An AC (or edge case) with no test covering it.
29	- A test marked `skip` / `xfail` / `it.skip` / `@pytest.mark.skip` / `t.Skip` / equivalent.
30	- A test with `TODO`, commented-out assertions, or empty body.
31	- An edge case from the spec without observable coverage.
32
33	Warnings (surface only, do not auto-fix):
34	- Tests that don't trace back to any AC or edge case in the spec. They might be intentional extra coverage or scope creep — that's the user's call. Listed in the report; do not affect the verdict.
35
36	5. Confirm green. Run the project's test command (from `.agents/tdd/test-command.txt` if it exists, otherwise infer it the same way the `tdd` skill does). All tests must pass.
37
38	6. Report and persist the verdict. Use the format below, and write the one-word verdict to `.agents/verify/last-verdict.txt` (see Persisting the verdict).
39
40	## Output format
41
42	```markdown
43	## Verify report — <slug>
44
45	### Acceptance criteria coverage
46	- ✅ AC1 — `tests/auth/test_token.py::test_expired_tokens_rejected`
47	- ✅ AC2 — `tests/auth/test_token.py::test_valid_tokens_accepted`
48	- ❌ AC3 — no test found
49
50	### Edge cases
51	- ✅ Empty input — `tests/auth/test_token.py::test_empty_token_returns_400`
52	- ❌ Concurrent refresh — no test
53
54	### Tests without spec mapping (warning — your call)
55	- ⚠️ `tests/auth/test_token.py::test_token_internal_repr` — does not map to any AC. Intentional extra coverage, or scope-creep test to delete?
56
57	### Test run
58	✅ 47 passed, 0 failed, 0 skipped.
59
60	### Verdict
61	Not ready. 2 items lack coverage. See gaps above.
62	```
63
64	When all blocking gaps are absent and the suite is green, the verdict is Ready for review. Tests-without-mapping warnings do not affect the verdict.
65
66	## Persisting the verdict
67
68	Every run of `verify` ends by writing the one-word verdict to `.agents/verify/last-verdict.txt` (create the directory if missing). The file is a single line, no trailing newline:
69
70	- `Ready` — all blocking gaps resolved and the suite is green.
71	- `Not ready` — at least one blocking gap remains.
72
73	The Stop hook reads this file. If you forget to write it, the previous run's verdict persists; if you never write it, the hook never blocks. Always overwrite at the end of every run.
74
75	## Where this skill ends
76
77	- Verdict = Ready for review → hand off to `review` (built-in) or proceed to PR.
78	- Verdict = Not ready → DO NOT END THE TURN. Immediately hand off to `tdd` with the gap list as the new test queue. After `tdd` is green, re-invoke `verify`. Repeat until the verdict is `Ready`.
79
80	The Stop hook forces this iteration: while the verdict file says `Not ready`, the turn cannot end. Do not try to skip the loop — fix the gaps via `tdd`, re-verify, repeat.
81
82	Do not try to fix gaps inside `verify` itself — that's `tdd`'s job. Verify

People who liked this also viewed

6 picks

Type

verify

15stars1forks

View on GitHub

How to install verify?

eduwxyz/my-awesome-skills/verify

$npx -y skills add eduwxyz/my-awesome-skills --skill verify

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Verify
2
3	The closing gate of the SDD pipeline. Map every acceptance criterion in the spec to a green test, surface any gaps, decide whether the work is ready for review.
4
5	## Skip verify for
6
7	- Spike / scratch work — there's no spec to verify against.
8	- Tiny obvious changes — typo fix, single-line bug, comment update.
9	- Specs without an `## Acceptance criteria` section — the upstream skill (`interview-to-spec` or `diagnose`) was abandoned mid-way. Finish that first.
10
11	## Before you start
12
13	Identify the spec. A path or slug for an existing `spec/<slug>.md`. If neither was provided, ask the user — do not guess.
14
15	Validate it. The spec must contain `## Acceptance criteria`. If missing, abort and tell the user to complete the spec first.
16
17	## The work
18
19	1. Read the spec. Focus on `Acceptance criteria` and `Edge cases` (or `Related risks` for bug specs).
20
21	2. Find the tests. Locate the relevant test files in the project's test directory (`tests/`, `__tests__/`, `_test.go`, `_spec.rb`, etc). Read them.
22
23	3. Map AC → test. For each acceptance criterion, find the test that observably proves it. Match on behaviour, not on naming similarity. A test named close to an AC does not count if it asserts something different.
24
25	4. Find gaps. Two categories — handled differently:
26
27	Blocking gaps (auto-fixed via handoff to `tdd`):
28	- An AC (or edge case) with no test covering it.
29	- A test marked `skip` / `xfail` / `it.skip` / `@pytest.mark.skip` / `t.Skip` / equivalent.
30	- A test with `TODO`, commented-out assertions, or empty body.
31	- An edge case from the spec without observable coverage.
32
33	Warnings (surface only, do not auto-fix):
34	- Tests that don't trace back to any AC or edge case in the spec. They might be intentional extra coverage or scope creep — that's the user's call. Listed in the report; do not affect the verdict.
35
36	5. Confirm green. Run the project's test command (from `.agents/tdd/test-command.txt` if it exists, otherwise infer it the same way the `tdd` skill does). All tests must pass.
37
38	6. Report and persist the verdict. Use the format below, and write the one-word verdict to `.agents/verify/last-verdict.txt` (see Persisting the verdict).
39
40	## Output format
41
42	```markdown
43	## Verify report — <slug>
44
45	### Acceptance criteria coverage
46	- ✅ AC1 — `tests/auth/test_token.py::test_expired_tokens_rejected`
47	- ✅ AC2 — `tests/auth/test_token.py::test_valid_tokens_accepted`
48	- ❌ AC3 — no test found
49
50	### Edge cases
51	- ✅ Empty input — `tests/auth/test_token.py::test_empty_token_returns_400`
52	- ❌ Concurrent refresh — no test
53
54	### Tests without spec mapping (warning — your call)
55	- ⚠️ `tests/auth/test_token.py::test_token_internal_repr` — does not map to any AC. Intentional extra coverage, or scope-creep test to delete?
56
57	### Test run
58	✅ 47 passed, 0 failed, 0 skipped.
59
60	### Verdict
61	Not ready. 2 items lack coverage. See gaps above.
62	```
63
64	When all blocking gaps are absent and the suite is green, the verdict is Ready for review. Tests-without-mapping warnings do not affect the verdict.
65
66	## Persisting the verdict
67
68	Every run of `verify` ends by writing the one-word verdict to `.agents/verify/last-verdict.txt` (create the directory if missing). The file is a single line, no trailing newline:
69
70	- `Ready` — all blocking gaps resolved and the suite is green.
71	- `Not ready` — at least one blocking gap remains.
72
73	The Stop hook reads this file. If you forget to write it, the previous run's verdict persists; if you never write it, the hook never blocks. Always overwrite at the end of every run.
74
75	## Where this skill ends
76
77	- Verdict = Ready for review → hand off to `review` (built-in) or proceed to PR.
78	- Verdict = Not ready → DO NOT END THE TURN. Immediately hand off to `tdd` with the gap list as the new test queue. After `tdd` is green, re-invoke `verify`. Repeat until the verdict is `Ready`.
79
80	The Stop hook forces this iteration: while the verdict file says `Not ready`, the turn cannot end. Do not try to skip the loop — fix the gaps via `tdd`, re-verify, repeat.
81
82	Do not try to fix gaps inside `verify` itself — that's `tdd`'s job. Verify

People who liked this also viewed

6 picks

Type