majiayu000 Documentation
Documentation-only workflow. Edit markdown files without touching application source code, configs, or infrastructure.
Scope
Allowed:
- Markdown docs, PRDs, ADRs
- Release notes
- Blog or marketing copy
- UI microcopy
Not allowed:
- Application source code
- Build config or infrastructure code
- Dependency manifests
1. Gather Context
Read:
- Root
AGENTS.md - Relevant scoped
AGENTS.md - Existing docs in the target area — understand structure, tone, terminology
- Related code (read-only) when needed for technical accuracy
2. Classify the Content
Identify the content type:
| Type | Approach |
|---|---|
| Internal technical docs | Concise, accurate, match existing style |
| Product or planning docs | Structured, stakeholder-readable |
| Public-facing (blog, landing page, marketing) | SEO-aware + humanizer pass |
| Release communication | Changelog format, audience-appropriate |
| UI microcopy | Concise, clear, action-oriented |
3. Write
Rules:
- English only unless the user requests otherwise
- Match existing terminology and conventions
- Keep examples technically accurate
- Prefer concise, scannable structure (Diataxis framework where appropriate)
- Progressive disclosure — overview first, details on demand
For Public Content
Additionally:
- Title tag: descriptive, matches search intent
- Meta description: compelling summary
- Heading hierarchy: one H1, logical H2 → H3 flow
- Internal links: descriptive anchor text to related content
- Images: descriptive alt text
- Structured data: Article/BlogPosting schema where applicable
Humanization (public content only)
Apply humanizer skill:
- Scan for AI writing patterns
- Rewrite problematic sections
- Anti-AI audit for text longer than 2 paragraphs
- Verify text sounds natural when read aloud
4. Verify
Check:
- All internal links resolve (no broken references)
- Code examples are accurate and match current implementation
- Terminology matches project conventions
- Formatting follows existing documentation patterns
- No secrets, PII, or internal-only information in public content
- No source code, config, or infrastructure files were changed
For Public Content — Humanization Checklist
- Text scanned for AI writing patterns
- Anti-AI audit performed for text longer than 2 paragraphs
- Text sounds natural when read aloud
For Public Content — SEO Checklist
- Title and meta description present and unique
- Heading hierarchy is logical (H1 → H2 → H3)
- Internal links with descriptive anchors added
- Images have alt text
- Page is indexable (no accidental noindex)
- Canonical URL is correct
5. Report
Summarize:
- Content type and Diataxis mode (Tutorial / How-to / Reference / Explanation)
- Files changed (created/updated — what changed)
- SEO review status (pass / N/A for internal docs)
- Humanization status (pass / N/A for internal docs)
- Verification result
- Next steps if any
Integration
- Skills:
humanizer(AI writing pattern removal for public-facing content),content-creation(content brief and tools for blog/landing pages) - Rules:
humanize-content(enforces humanizer pass on public content) - Follow-up:
seo-review(for public content),pre-commit,create-pr