homebrew-tools

A Homebrew tap with my CLI utilities. Right now it ships four linters for Claude Code and MCP configuration, packaged as self-contained Homebrew formulae so you don't need to manage a Python environment yourself.

Tap it

brew tap mukundakatta/tools

Available formulae

Formula	What it does
claude-skill-check	Lint Claude Code SKILL.md files (YAML frontmatter, required fields, description length, leaked secrets).
mcp-config-check	Lint MCP (Model Context Protocol) config files for Claude Desktop, Cursor, Cline, Windsurf, Zed.
claude-hooks-check	Lint the hooks block of a Claude Code settings.json (typos, dangerous commands, leaked secrets).
claude-commands-check	Lint Claude Code slash-command files in .claude/commands/.
llm-usage-report	Parse LLM API response logs (Anthropic, OpenAI, Google) and generate token + cost reports.
ai-eval-forge	Zero-dependency eval harness for LLM and agent regression testing. Scores outputs with exact, contains, regex, token-F1, JSON, and citation checks.
agent-run-diff	Compare baseline vs current agent runs; surface regressions as structured reasons across 7 signals (success loss, new errors, tool failures, output drift, step/latency/cost bloat).
streamparse	Streaming JSON parser CLI. parse, extract, validate subcommands for partial / messy / truncated JSON from LLM streams.
streamparse-mcp	MCP server exposing streamparse to Claude Desktop, Cursor, Cline, Windsurf, and Zed. Lets the assistant parse partial JSON on demand.

Install one (or more)

brew install claude-skill-check
brew install mcp-config-check
brew install claude-hooks-check
brew install claude-commands-check

Or install all four at once:

brew install claude-skill-check mcp-config-check claude-hooks-check claude-commands-check

Use

Each formula installs a CLI of the same name:

claude-skill-check path/to/SKILL.md
mcp-config-check ~/Library/Application\ Support/Claude/claude_desktop_config.json
claude-hooks-check ~/.claude/settings.json
claude-commands-check .claude/commands/

Exit status is 0 on no errors and 1 on any errors, so the CLIs compose cleanly in git hooks and CI.

Source and upstream releases

Python formulae pull their sdist from PyPI. Node-based formulae pull tarballs from npm.

Python (PyPI)

• claude-skill-check — also on PyPI
• mcp-config-check — also on PyPI
• claude-hooks-check — also on PyPI
• claude-commands-check — also on PyPI
• llm-usage-report — also on PyPI
• ai-eval-forge — also on PyPI
• agent-run-diff — also on PyPI

Node (npm)

• streamparse — also on npm
• streamparse-mcp — also on npm

Updating

brew update
brew upgrade claude-skill-check   # or any of the others

Uninstall the tap

brew untap mukundakatta/tools

License

Each formula carries the upstream project's license (MIT across all four). This tap itself is MIT-licensed.