Context7 Documentation Skill

Fetch version-specific, up-to-date documentation for any library or framework using the Context7 MCP server. This skill is used proactively — you do not need to wait for the user to ask for docs.

When to use this skill

Use Context7 automatically whenever you:

• •Need to write code using a library's API
• •Are generating setup or configuration steps
• •Need to verify correct function signatures, options, or patterns
• •Are troubleshooting library-related errors
• •Are unsure about current API conventions (your training data may be outdated)
• •Are working with any dependency listed in the project's package manager files

Do NOT use for general programming concepts, language builtins, or topics unrelated to specific libraries.

Instructions

Step 1: Check the registry

Before calling Context7, read .claude/context7-registry.md to see if the library already has a known Context7 identifier.

If the library is listed with a context7_id, skip to Step 3.

Step 2: Resolve the library ID

If the library is not in the registry, use the Context7 MCP tool to resolve it:

Tool: mcp__context7__resolve-library-id
Input: { "libraryName": "<library-name>" }

Tips for resolving:

• •Use the npm/PyPI/crates.io package name (e.g., react, express, tauri)
• •If the first attempt fails, try the full name (e.g., @tauri-apps/api instead of tauri)
• •For monorepo packages, try both the scope and the individual package

Once resolved, immediately register the library — see Step 5.

Step 3: Fetch documentation

Use the Context7 MCP tool to get documentation:

Tool: mcp__context7__get-library-docs
Input: {
  "context7CompatibleLibraryID": "<context7_id from registry>",
  "topic": "<specific topic you need>"
}

Topic tips:

• •Be specific: "useEffect hook" not "hooks"
• •Include the operation: "file upload configuration" not "files"
• •For setup: "getting started" or "installation"

Step 4: Apply the documentation

• •Use the fetched docs to write correct, up-to-date code
• •Prefer patterns shown in the official docs over patterns from training data
• •If the docs show a different API than expected, trust the docs

Step 5: Update the registry

After resolving a new library or fetching docs, update .claude/context7-registry.md:

1. •Read the current registry file
2. •Add or update the entry with:
   • •Library name (package manager name)
   • •Context7 ID (from resolve step)
   • •Version currently in use (from package.json, Cargo.toml, etc.)
   • •Date last verified
3. •Write the updated registry

Registry entry format:

| library-name | /org/repo | 1.2.3 | 2025-01-15 |

Step 6: Version checking (periodic)

At the start of each session and whenever you read a package manager file (package.json, Cargo.toml, etc.):

1. •Read the registry at .claude/context7-registry.md
2. •Read the project's dependency file(s)
3. •Compare versions for all registered libraries
4. •For any mismatch:
   • •Warn the user: "Library X version changed: registry says 1.2.3, project uses 2.0.0"
   • •Auto-update the registry with the new version and today's date
   • •Re-fetch docs if the major version changed (API may have breaking changes)

Error handling

• •Context7 MCP not available: Inform the user that the Context7 plugin needs to be enabled. They can enable it via /install-plugin context7 or by adding "context7@context7": true to enabledPlugins in ~/.claude/settings.json.
• •Library not found: Try alternative names (scoped, unscoped, org name). If still not found, note it in the registry as NOT_FOUND so we don't retry every time.
• •Rate limited or timeout: Fall back to your training data but note that the docs may be outdated.

Registry file location

The registry lives at: .claude/context7-registry.md

This file is committed to the repo so all team members share the same library mappings.