# Cursor Docs

> Fetch external/official documentation for a library, API, or framework and cache it locally as clean Markdown with a manifest, so it can be referenced authoritatively without re-fetching. Use this whenever the user wants to "index", "add", "pull in", or "cache" docs, references a specific library/API and wants accurate guidance, or says "@docs" / "use the docs for X". Also consult the local cache before answering questions about a library that has cached docs. Invokable as /cursor-docs.

- Skill: `jambot24/cursor-docs` (Agent Skill)
- Install (CLI): `npx skillmds@latest add jambot24/cursor-docs`
- Raw SKILL.md: https://api.skillmd.com/api/skills/jambot24/cursor-docs/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Docs & Writing
- Author: jambot24 (https://skillmd.com/u/jambot24)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/jambot24/cursor-docs

---


# Cursor Docs

A local, versioned analog to Cursor's `@Docs`. Instead of a server-side crawler
and vector index, this caches authoritative documentation as Markdown in the
repo, where it is reviewable, diffable, and free to re-read. This mirrors the
"download official docs as Markdown locally" pattern, which is more reliable
than live crawling.

## Storage layout

```
docs/
  <source-slug>/
    _index.md        # manifest: source URLs, fetch date, version
    <page>.md        # one file per page/section
```

`<source-slug>` is a short kebab-case name, e.g. `microsoft-graph`,
`ninjaone-api`, `stripe`.

## Add or update docs

1. **Resolve the source.** If given a library name rather than a URL, use
   `WebSearch` to find the official documentation, prefer first-party sources
   (vendor docs, Microsoft Learn, RFCs) over aggregators.
2. **Fetch.** Use `WebFetch` on the target page(s). For multi-page docs, fetch
   the specific pages relevant to the task rather than the entire site.
3. **Clean.** Strip navigation, footers, cookie banners, and repeated
   boilerplate. Keep headings, prose, code blocks, and tables. Preserve the
   source's own structure.
4. **Save.** Write each page to `docs/<source-slug>/<page>.md`.
5. **Record the manifest.** Create or update `docs/<source-slug>/_index.md`:

   ```markdown
   # <Source name> docs cache

   | Page | Source URL | Fetched (ISO 8601) | Version |
   |------|-----------|--------------------|---------|
   | overview.md | https://... | 2026-06-15 | v1.2 |
   ```

   Record the documented version/date if the source states one.

## Reference docs

When a task involves a library that has a cache under `docs/`, read the relevant
`docs/<source-slug>/*.md` files first and base guidance on them. Quote the
cached source rather than relying on memory. If the cache does not cover the
needed topic, fetch and add it (workflow above) before answering.

## Staleness

The manifest's fetch date is the source of truth for freshness. If a cached
source is older than ~90 days, or the user reports the docs changed, offer to
re-fetch and update the manifest. Never silently trust a stale cache for fast-
moving APIs without noting its age.

## Rules
- First-party sources only where possible; note when a source is unofficial.
- Never fabricate API details, if it is not in the cache or fetchable, say so.
- Keep cached files clean enough to review in a PR.