菜单
Lint Cadenza spec files and surface structural findings. Use this when the user invokes cadenza-spec-lint, asks to validate spec structure, edits spec/**/*.md, or needs the spec:lint completion gate.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Best-practice authoring guidance for Cadenza decks. Use this skill whenever the user asks to create, revise, debug, validate, or review a Cadenza deck, Cadenza TSX, Cadenza skills, render-safe components, typed slide APIs, speaker notes, motion timing, browser preview behavior, or diagnostics-driven repair, even if they do not explicitly say "skill" or "best practices."
Run Cadenza Builder implementation work across phases. Use this skill whenever the user asks for Builder, a SPEC_TEST_MATRIX batch, TDD, RED GREEN REFACTOR, remediation implementation, trace update, or package/test work in Cadenza, even when they only name a phase kick file. This skill requires the phase kick file and frozen specs for concrete scope.
Run Cadenza Architect work across phases. Use this skill whenever the user asks for Architect, Stage A, Stage B, specs, Freeze Candidates, contract design, ADR routing, or phase contract freeze work in Cadenza, even when they only name a phase kick file. This skill preserves phase-specific kick files as the concrete task entrypoints.
Prepare Cadenza Wizard handoffs at workflow boundaries. Use this skill whenever the user asks for Wizard work, a next Architect kick file, phase-boundary handoff, Phase 0 bootstrap from Scout to Architect, Scout-to-Architect planning, or any Cadenza handoff that must summarize accepted inputs without changing specs, production code, or phase pointers.
Run an independent Cadenza review after Builder work, phase closeout, or maintainer-selected high-risk changes. Use this skill whenever the user asks for reviewer, review Builder output, audit implementation against frozen specs, validate phase closeout, or produce remediation findings for Cadenza, even if they do not explicitly name the skill.
Run Cadenza's cold-start onboarding gate. Use this when the user explicitly invokes cadenza-onboard, asks to onboard a fresh Cadenza agent session, resumes after compaction, or needs the AGENTS.md startup protocol before doing project work.
| name | cadenza-spec-lint |
| description | Lint Cadenza spec files and surface structural findings. Use this when the user invokes cadenza-spec-lint, asks to validate spec structure, edits spec/**/*.md, or needs the spec:lint completion gate. |
Use this skill to validate Cadenza spec/ Markdown contracts. Treat it as a
read-only verification workflow unless the user separately asks you to fix
findings.
If the user passed an argument, treat it as the target scope. Otherwise lint the
whole spec/ tree.
If package.json exists and defines a spec:lint script, run:
pnpm spec:lint <target-if-any>
Capture and report the exit code.
If package.json is absent or has no spec:lint script, report:
Phase 0 pre-B-M4: `pnpm spec:lint` is not yet implemented. Falling back to bundled hygiene.
Then run the bundled fallback script from this skill directory:
bash ${CLAUDE_SKILL_DIR}/scripts/spec-lint-fallback.sh <target-if-any>
In non-Claude runtimes, locate the active cadenza-spec-lint skill
directory and run scripts/spec-lint-fallback.sh from there.
Return this structure:
Scope: <file or "all specs">
Tool used: <pnpm spec:lint | bundled fallback>
Files scanned: <N>
Findings:
- <file> : <level: info|warn|error> : <message>
Summary: <OK | N warnings | N errors>
Exit: <0 or non-zero>
If errors are present, do not claim the work is done. Cadenza completion gates require verification commands to exit successfully.