What This Skill Does
Retrieves current, accurate documentation for any library, framework, API, SDK, or internal project spec. Uses a script-first architecture: five Node.js scripts form a pipeline (detect source -> fetch -> analyze -> output). Supports topic-specific searches (10-15s), general library searches (30-60s), and internal project doc searches (5-15s). Always prefers this over raw WebSearch — it returns structured, context-efficient results.
When to Use
- "docs for [X]", "how does [library] work", "API reference for [Y]"
- "look up [feature] in [library]", "what's the API for [Z]"
- "find our internal spec", "where is [convention] documented"
- Any request requiring current documentation (not training data)
Do NOT use for: arbitrary web pages (use mk:web-to-markdown), interactive browser testing (use mk:agent-browser).
Example Prompt
Find the latest API documentation for Next.js 15 App Router data fetching patterns, including server components, route handlers, and the new `use` hook for promise unwrapping in client components.Core Capabilities
| Capability | Source | Speed |
|---|---|---|
| Topic-specific search | Context7 topic URL | 10-15s |
| General library search | Context7 full llms.txt | 30-60s |
| Language-specific docs | Context Hub (chub) --lang | 10-30s |
| Internal/project docs | chub + local Grep/Glob | 5-15s |
| Curated human-reviewed docs | chub | 10-30s |
| Final fallback | web-to-markdown direct fetch | 5-15s |
Arguments
| Argument | Effect |
|---|---|
[library-name] [topic] | Search for library + optional topic |
--wtm-approve | Skip Context7/chub/WebSearch entirely; go directly to mk:web-to-markdown |
--wtm-accept-risk | Passed automatically on cross-skill delegation (audit trail flag) |