菜单
Validate canonical review-doc source and rendered HTML across layers — JSON Schema, Mermaid syntax/render, accessibility (Axe/WCAG AA), responsive layout, and print/PDF. Use before sharing or committing a branded review artifact, in CI, or as a pre-commit hook.
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。
Author compact canonical source for an ARCHITECTURE review (artifact_type architecture-review) — decisions, constraints, interfaces, deployment context, alternatives, and interaction diagrams. Produces content-only YAML/JSON validated against the base + architecture overlay schema; render with render-branded-html. Use for architecture/engineering reviews.
Author compact canonical source for a PRD review (artifact_type prd-review) — business goals, out-of-scope, requirement coverage, stakeholders, readiness. Produces content-only YAML/JSON validated against the base + PRD overlay schema; render with render-branded-html. Use for product/delivery PRD reviews.
Deterministically render validated canonical review-doc source (YAML/JSON) into a single self-contained branded HTML file, with three layout skins (docs-site, editorial, brief), light/dark/auto mode, a widget section library, interactive or frozen Mermaid with fullscreen zoom, and optional PDF. Use when a review artifact's canonical source is ready and you need the styled HTML/PDF output. The model authors content; this renderer owns all chrome.
Entry point for producing a branded review document. Classifies the request into one of five artifact types (user-story, architecture, PRD, risk-analysis, test-charter), gathers evidence scope, delegates content authoring to the matching artifact skill, and renders/validates the result. Use when someone asks for a reviewable, company-themed doc from analysis (e.g. "make an architecture review", "turn this into a risk register HTML").
Author compact canonical source for a RISK ANALYSIS review (artifact_type risk-analysis) — risk matrix, triggers, controls, residual risk, owners, review cadence. Produces content-only YAML/JSON validated against the base + risk overlay schema; render with render-branded-html. Use for QA/product/operations risk reviews.
Author compact canonical source for an exploratory TEST CHARTER (artifact_type test-charter) — missions, heuristics, coverage boundaries, exit criteria, states explored. Produces content-only YAML/JSON validated against the base + test-charter overlay schema; render with render-branded-html. Use for QA exploratory-testing charters.
| name | validate-branded-doc |
| description | Validate canonical review-doc source and rendered HTML across layers — JSON Schema, Mermaid syntax/render, accessibility (Axe/WCAG AA), responsive layout, and print/PDF. Use before sharing or committing a branded review artifact, in CI, or as a pre-commit hook. |
Layered validation for branded review documents. Schema validation runs first and short-circuits the rest on failure — a structurally invalid artifact never reaches the Mermaid, accessibility, responsive, or export checks.
| Layer | Script | Pass condition |
|---|---|---|
| Schema | scripts/validate_schema.py | all required fields + enums valid (base + overlay) |
| Mermaid | scripts/validate_mermaid.mjs | every diagram parses (add --render for mmdc render in CI) |
| Accessibility | scripts/validate_accessibility.mjs | zero critical Axe violations; contrast ≥ WCAG AA |
| Responsive | scripts/validate_responsive.mjs | no horizontal overflow at desktop + mobile |
| Print/PDF | scripts/validate_export.mjs | PDF exports with ≥1 readable page; print completeness — every tab pane and details body visible, screen-only controls hidden |
For sources declaring theme.mode: dark or auto, an additional
accessibility-dark pass runs under dark color-scheme emulation
(--color-scheme dark) with the same WCAG AA thresholds.
The accessibility layer also runs widget checks: tabs keyboard-operable with associated labels, diagram zoom toggles carry accessible names, and severity / callout tag contrast is enforced via Axe's WCAG AA color-contrast rule.
Run the whole suite (renders HTML automatically if --html is omitted):
python .claude/skills/validate-branded-doc/scripts/validate_all.py \
--input examples/architecture-review.yaml
Validate the same source under an explicit skin (matrix testing in CI):
python .claude/skills/validate-branded-doc/scripts/validate_all.py \
--input examples/architecture-review.yaml --skin editorial
Run a single layer:
python .claude/skills/validate-branded-doc/scripts/validate_schema.py --input source.yaml
node .claude/skills/validate-branded-doc/scripts/validate_mermaid.mjs --input source.yaml --render
node .claude/skills/validate-branded-doc/scripts/validate_accessibility.mjs --input doc.html
SKIP and fails the run.playwright,
@axe-core/playwright). When absent they report SKIP (not FAIL) so the
schema + Mermaid gate still works offline. Install deps to enforce them — see
the renderer skill's setup notes.0 all pass, 2 any explicit failure. SKIP never fails the run.--json for machine-readable reports (CI)..venv): jsonschema, pyyaml, referencing.playwright, @axe-core/playwright; optional @mermaid-js/mermaid-cli for --render.