القائمة
Runs scripts/bench.ts against the performance budget and reports the result. Fails if any scenario exceeds budget. Appends a row to docs/performance.md on success.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
Parallel-track dispatcher. Reads .claude/backlog.md, picks the next unchecked item from each of up to 3 active tracks, dispatches them in parallel to specialist subagents, verifies, commits, and checks off. The primary entry point for autonomous ra11y development.
Scaffolds a new accessibility standard module (metadata, criteria, standard export, tests, kb entry). Delegates the criteria enumeration to spec-researcher + standard-builder. Use when adding any conformance framework.
Scaffolds a new ra11y rule end-to-end from a criterion ID using the evaluator-optimizer pattern — spec-researcher produces WCAG context, rule-implementer generates, a11y-reviewer critiques, iterate up to 3 times, then code-reviewer signs off. Use when adding any new rule.
Universal preflight — runs the full check sequence (typecheck, lint, tests, zero-dep, cycles, network isolation, limits, tsdoc, mermaid, kb drift) through the single `bun run verify` entrypoint. Use before every commit and before handing off to another agent.
Scaffolds a new output formatter — file in src/output/formatters/, snapshot tests, CLI --format registration, and a short doc stub. Use when adding terminal/json/sarif/junit/html/markdown/plain or a custom format.
Regenerates auto-generated docs (docs/kb/wcag/, docs/kb/rules/, kb index, llms.txt, docs/api/) and commits the result so check-kb-drift passes. Use after any rule or standard change.
استنادا إلى تصنيف SOC المهني
| name | bench |
| description | Runs scripts/bench.ts against the performance budget and reports the result. Fails if any scenario exceeds budget. Appends a row to docs/performance.md on success. |
| allowed-tools | Bash(bun *) Bash(git *) |
Runs the benchmark harness and checks against the budget in CLAUDE.md section 13.
bun run build — bench runs against the built artifact, not source.bun scripts/bench.ts --json — capture structured results.docs/performance.md with date | commit | scenario | ms | budget | margin, stage, and commit: perf(bench): record <shortsha> results.benchmark-tuner with the failing scenario and its measurement.scenarios:
cold_start: <n>ms (budget 200ms, margin +<x>ms)
10_files: <n>ms (budget 100ms)
100_files: <n>ms (budget 500ms)
1000_files: <n>ms (budget 3000ms)
decision: PASS | FAIL
recorded: docs/performance.md:<line>