Skip to content

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

CapabilitySourceSpeed
Topic-specific searchContext7 topic URL10-15s
General library searchContext7 full llms.txt30-60s
Language-specific docsContext Hub (chub) --lang10-30s
Internal/project docschub + local Grep/Glob5-15s
Curated human-reviewed docschub10-30s
Final fallbackweb-to-markdown direct fetch5-15s

Arguments

ArgumentEffect
[library-name] [topic]Search for library + optional topic
--wtm-approveSkip Context7/chub/WebSearch entirely; go directly to mk:web-to-markdown
--wtm-accept-riskPassed automatically on cross-skill delegation (audit trail flag)

Workflow

Tiered Fallback Chain (default)

Released under the MIT License.