Menu
Tag tests with AC references so /df:verify L3 can confirm every spec AC is exercised by at least one test
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
Encapsulates the DECISIONS: tagging protocol for architectural decision documentation in task outputs
Produces structured JSON intelligence for a remote GitHub repo — fetches metadata and file tree via gh api, reads key files via WebFetch. No local clone. Use when evaluating an unfamiliar repo before planning integration work.
Skeleton skill for eval fixture — replace with the real skill content
Fetches live web content using headless Chromium via Playwright. Use when you need to read documentation, articles, or any public URL that requires JavaScript rendering. Falls back to WebFetch for simple HTML pages.
Verifies UI acceptance criteria by launching a headless browser, extracting the accessibility tree, and evaluating structured assertions deterministically. Use when a spec has browser-based ACs that need automated verification after implementation.
Discovers requirement gaps during ideation. Use when user describes features, planning specs, or requirements seem incomplete. Asks clarifying questions about scope, constraints, edge cases, success criteria.
| name | df-ac-coverage |
| description | Tag tests with AC references so /df:verify L3 can confirm every spec AC is exercised by at least one test |
/df:verify L3 is a lint, not a self-report. It checks that every AC in a spec is referenced by at least one test file (or explicitly marked [advisory]). Whether the test passes is L4's job — L3 only checks the tag exists.
In any test file (test name, comment, JSDoc, string literal — anywhere in the file), include the literal string:
specs/<slug>.md#AC-<n>
Where <slug> is the spec basename without the doing-/done- prefix and .md suffix.
// covers specs/upload.md#AC-1
test('rejects non-image MIME types', () => { /* ... */ });
/**
* @covers specs/upload.md#AC-3
*/
function helper() {}
test('specs/upload.md#AC-5: error boundary catches render failures', () => {});
All three are valid — the regex looks for the literal specs/<slug>.md#AC-N substring anywhere in the file.
Some ACs cannot be machine-verified (e.g. "code is readable" or "comments explain rationale"). Mark these in the spec by adding [advisory] to the AC bullet:
- [ ] **AC-7** — [advisory] WHEN reviewers read the diff THEN the rationale comments SHALL be clear.
L3 skips advisory ACs.
hooks/ac-coverage.js (CLI mode) is invoked by /df:verify L3:
node "${HOME}/.claude/hooks/ac-coverage.js" --spec {spec_path} --snapshot .deepflow/auto-snapshot.txt --status pass
It also runs as a PostToolUse hook on git commit to catch drift mid-flight. Exit 2 with OVERRIDE:SALVAGEABLE lists untagged non-advisory ACs.
Earlier versions asked agents to emit AC_COVERAGE: … AC_COVERAGE_END blocks in their output, which execute.md parsed into .deepflow/results/T<n>.yaml for L3 to read back. That was the agent grading itself — no real signal. This skill replaces the entire self-report path with a simple test-tag presence check.