Skip to content

Tools Reference

Python tools that skills invoke via Bash. Claude calls these automatically — you don't run them directly unless troubleshooting.

tools/read-section/read_section.py

Extracts one function, class, or section from a source file without loading the rest.

python tools/read-section/read_section.py --file src/auth.ts --fn validateToken
python tools/read-section/read_section.py --file docs/guide.md --after "## Install"
python tools/read-section/read_section.py --file handler.go --lines 50-80

Language detection is automatic. Supports: Python (indent-based), JS/TS/Go/Java/C#/Rust (brace-counting), Markdown (heading hierarchy), others (brace → indent fallback).

Exit code 0 = found; exit code 1 = not found or unreadable.


tools/env-check/env_check.py

Checks runtime environment: language versions, open ports, Docker container status, .env file presence and git-ignore status.

python tools/env-check/env_check.py

Output is structured JSON. The /env-check skill parses it and returns a summary.


tools/pkg-info/pkg_info.py

Queries npm or PyPI for package metadata.

python tools/pkg-info/pkg_info.py --pkg drizzle-orm --registry npm
python tools/pkg-info/pkg_info.py --pkg fastapi --registry pypi

Returns: latest version, TypeScript types, license, weekly downloads, homepage.


tools/diff-summary/diff_summary.py

Produces a structured summary of git diff output — file-by-file with a one-line description per change.

python tools/diff-summary/diff_summary.py

Reads from git diff on stdin or runs git diff HEAD automatically.


tools/git-ctx/git_ctx.py

Batches multiple git queries into one structured output: branch, status, recent log, and diff stat.

python tools/git-ctx/git_ctx.py

tools/indexer/indexer.py

Builds the structural code index for the project. Run once via /index-repo.

python tools/indexer/indexer.py --project-dir /path/to/project --output .claude/index/

Produces: files.json, symbols.json, graph-imports.json, graph-calls.json, graph-clusters.json, manifest.json. Supports incremental re-runs via the manifest.


tools/crawl4ai/fetch.py

Fetches a URL through the local Crawl4AI container and returns clean Markdown.

python tools/crawl4ai/fetch.py --url "https://example.com/docs"

Caches output at vault/07-web-cache/ for 24 hours. Falls back to printing "not reachable" on stderr if Crawl4AI is down (callers fall back to built-in WebFetch).