​
Overview

Treat documentation as a product. Use Blink to plan structure, draft content, add citations, and keep docs synchronized with code. Write for two audiences at once: humans and AI assistants that will ingest your docs.

​
Before You Start

  • Structure content with consistent headings, summaries, and self-contained sections.
  • Write clear instructions, examples, and explicit constraints over prose.
  • Optimize for chunkability and retrieval: avoid hidden dependencies between sections.
  • Include troubleshooting and FAQs; they are high-signal for both users and LLMs.

​
Workflow

  1. Define audience, outcomes, and constraints
  2. Draft an outline (page tree, slugs, summaries)
  3. Generate content with examples, code, and callouts
  4. Add citations and cross-links; verify with scenarios
  5. Ship via docs-as-code and set owners + review cadence

​
Prompt Templates

Create a docs outline for "{feature}":
- Audience: {developer / admin / end-user}
- Goals: {what success looks like}
- Constraints: {platforms, versions, security}
- Required sections: Overview, Setup, Examples, Troubleshooting, FAQs
- Output: page tree with slugs and 1–2 sentence summaries

Draft a self-contained page for "{topic}" following our style:
- Start with a 2–3 sentence summary and prerequisites
- Provide numbered steps with exact commands/flags
- Include minimal, copyable examples for {language}
- Add a Troubleshooting section (3–5 issues with causes/fixes)
- End with a Verification checklist

Refactor existing docs to be AI-friendly:
- Break long pages into sections (≤ 300–500 words each)
- Make each section answerable standalone (no implicit context)
- Add explicit cross-links for dependencies
- Normalize terminology and headings across pages

Explore and document API reference and routes:
- Inspect router and handlers to list endpoints, methods, and auth
- For each route include: path, method, params, example request/response, errors
- Map internal modules (e.g., files, chat, webhooks) to docs pages/slugs
- Add notes on rate limits, pagination, and webhooks
- Output: REST API page skeleton + checklist of missing examples

​
Anti‑Patterns and Recovery

  • Wall-of-text pages without structure → add headings, summaries, and examples
  • Missing essential journeys (API keys, auth, deletion) → document core tasks first
  • Hidden assumptions between sections → make dependencies explicit and add links
  • Out-of-sync docs with code → add prompts to generate diffs and update notes
  • Structured outlines with page hierarchies and cross-links
  • Self-contained sections that work for humans and AI retrieval
  • Code examples and troubleshooting matched to your stack