fetch-docs

Fetch Docs

Retrieve up-to-date documentation for any library or framework and save it to the project knowledge base. Future conversations can read cached docs instead of re-fetching.

Usage

/fetch-docs next.js app-router          # Fetch Next.js App Router docs
/fetch-docs capacitor geolocation       # Fetch Capacitor Geolocation plugin docs
/fetch-docs leaflet markers             # Fetch Leaflet marker docs
/fetch-docs zustand persist-middleware  # Fetch Zustand persist docs

Auto-Trigger Rules

This skill's persistence behavior applies ANY TIME documentation enters the conversation, not just when explicitly invoked. This includes:

1. AI-fetched: Context7 lookups, WebFetch, WebSearch for API references or guides
2. User-provided: User pastes documentation, API specs, READMEs, or guides into chat
3. User-linked: User shares a URL to docs — fetch it and persist
4. User-shared files: User asks you to read a doc/PDF/spec file — extract and persist
5. Error solutions: When investigating errors leads to documentation or patterns worth keeping
6. Offline/unreachable docs: User provides docs because the AI can't reach them — these are especially valuable to persist since they can't be re-fetched

You MUST save the useful portions to the knowledge base before continuing with the task.

Handling user-provided documentation

When a user pastes or shares documentation that Claude cannot normally access (internal APIs, paywalled docs, private specs):

1. Acknowledge that this is being saved to the knowledge base
2. Structure it into the standard reference format (tables, code examples, key notes)
3. Tag it with <!-- user-provided, not auto-fetchable --> so future updates know to ask the user rather than trying to fetch
4. Thank briefly — the user is investing time to improve the knowledge base

Process

1. Check Cache First

Before fetching, check if docs already exist:

.claude/skills/fetch-docs/references/{library-name}.md