Docs Seeker

Find authoritative, up-to-date technical documentation for any library or framework before implementing.

What This Skill Does

The Challenge: AI models have training cutoffs. APIs change, libraries release breaking updates, and deprecated patterns persist in training data. Implementing based on stale knowledge produces code that doesn’t work with current versions.

The Solution: Docs Seeker skill queries live documentation sources via llms.txt endpoints and the Context7 MCP server. Returns current API references, migration guides, and code examples from official sources — always reflecting the latest stable release.

Activation

Implicit: Activates when user asks “how do I use X”, “what’s the API for Y”, or requests implementation guidance for a specific library.

Explicit: Activate via prompt:

Activate docs-seeker skill to find docs for [library/framework/version]

Capabilities

1. llms.txt Documentation Sources

Query official machine-readable documentation.

Supported format: Sites exposing /llms.txt or /llms-full.txt provide structured docs optimized for LLM consumption.

Query pattern:

Use docs-seeker to find the latest [library] documentation for [specific API/feature]

Popular sources available:

  • Next.js, Astro, Nuxt (frontend frameworks)
  • Prisma, Drizzle (ORMs)
  • Better Auth, Clerk (authentication)
  • Tailwind CSS, Radix UI (styling/components)
  • Cloudflare Workers, Vercel (deployment)

2. Context7 MCP Integration

Fetch library documentation programmatically via MCP server.

Context7 usage:

Use context7 to get docs for:
- Library: react-query
- Version: v5
- Topic: useQuery hook options

Returns: Current API signatures, parameter descriptions, return types, and official examples.

3. Version-Specific Lookup

Retrieve docs for a specific library version.

Example queries:

  • “Next.js 15 App Router route handlers”
  • “Prisma v6 migration workflow”
  • “Better Auth v1 OAuth configuration”

4. Cross-Reference Validation

Compare official docs against implementation to catch outdated patterns.

Validation workflow:

  1. Fetch current official example for the pattern in question
  2. Compare with implementation
  3. Flag deprecated APIs or changed method signatures

Prerequisites

For Context7 MCP:

  • MCP server configured in .claude/mcp.json
  • Internet access from Claude Code session

For llms.txt:

  • Library must expose /llms.txt endpoint
  • No additional setup required

Best Practices

1. Always check docs before implementing unfamiliar APIs A 30-second doc lookup prevents hours of debugging deprecated patterns.

2. Note the library version Specify exact version when searching. “Prisma” and “Prisma v6” return different answers.

3. Cross-reference with changelog For major version jumps, also check the migration guide.

Common Use Cases

Use Case 1: Implementing New Auth Library

Scenario: Setting up Better Auth for the first time.

Workflow:

  1. Activate docs-seeker for “Better Auth latest”
  2. Fetch installation, server setup, client setup sections
  3. Look up specific provider (Google OAuth) configuration
  4. Implement following official examples, not training data

Use Case 2: Upgrading Framework Version

Scenario: Upgrading Next.js from 14 to 15.

Workflow:

  1. Fetch Next.js 15 migration guide
  2. Identify breaking changes relevant to current codebase
  3. Fetch new App Router APIs replaced old patterns
  4. Plan upgrade phase by phase

Troubleshooting

Issue: Library not found in llms.txt sources Solution: Use Context7 MCP for broader coverage. Fall back to WebFetch from official docs URL.

Issue: Docs returned are for wrong version Solution: Specify version explicitly in query. Check if docs URL includes version in path.

  • /ckm:ask - General technical questions
  • /ckm:brainstorm - Evaluate library options