Documentation

Safe documentation-only workflow. Edit markdown files without touching application source code, configs, or infrastructure. Applies Agent(content-writer) for all writing tasks.

**⚠️ CONSTRAINT: This workflow NEVER modifies source code (*.java, *.ts, *.tsx, .py, .go), configs (.yaml, .yml, .json), infrastructure (.tf, Dockerfile, Helm), or dependency files (pom.xml, package.json, requirements.txt).

1. Define Scope

Ask the user (or extract from parent workflow):

• What documentation? (technical docs, API reference, README, PRD, design doc, ADR, blog post, release notes, UI copy)
• Content type:
  • Internal (technical docs, design docs, ADRs) → Agent(content-writer) only
  • Public-facing (blog, landing page, marketing) → Agent(content-writer) + Agent(seo-engineer)
• Action: Create new / update existing / restructure
• Target files: Which .md files will be affected

2. Apply Roles

3. Gather Context

Before writing:

1. Read project's CLAUDE.md — terminology, conventions, tech stack
2. Read existing docs — understand structure, tone, terminology already in use
3. Identify Diátaxis mode — Tutorial, How-to, Reference, or Explanation
4. Check related code (read-only) — verify technical accuracy of claims

4. Write Content

Follow Agent(content-writer) standards:

• Diátaxis framework for documentation structure
• English only unless explicitly requested otherwise
• Progressive disclosure — overview first, details on demand
• Tested examples — all code snippets must be accurate
• Consistent terminology — match existing project conventions

For Blog / Public Content

Additionally apply @humanizer skill — scan for and remove AI writing patterns. Then follow Agent(seo-engineer) standards:

• 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 (JSON-LD) where applicable
• No keyword stuffing — write for users, not crawlers

5. Verify

• All internal links are valid (no broken references)
• Code examples are accurate and match current implementation
• Terminology is consistent with 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 modified

For Public Content — Humanization Checklist

• Text scanned for AI writing patterns (@humanizer skill)
• 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
• Structured data validates (Rich Results Test)

6. Summary

## Documentation Summary
- **Type**: [technical docs / blog / API reference / release notes / etc.]
- **Content mode**: [Tutorial / How-to / Reference / Explanation]
- **Files changed**:
  - [file1.md]: [created/updated — what changed]
  - [file2.md]: [created/updated — what changed]
- **Role(s) applied**: [Agent(content-writer), Agent(seo-engineer) if public]
- **SEO review**: [pass / N/A for internal docs]
- **Verification**: [links valid, formatting correct, no code files modified]
- **Next steps**: [if any]

Integration

• Roles: Agent(content-writer) (primary), Agent(seo-engineer) (public-facing content), Agent(product-manager) (PRDs)
• Skills: @humanizer (AI writing pattern removal for public-facing content)
• Rules: humanize-content (auto-enforces humanizer pass)
• Follow-up: /seo-review (for public content), /pre-commit, /create-pr
• Related: /feature-dev (inline docs during development), /release (release notes)