Show HN: Guard hooks for Claude Code, tested against a public corpus

Category: security

Tags: claude-code, security, guardrails, ai-agent, developer-tools

Score: 7.3/10 (Innovation: 6, Technical: 7, Documentation: 9, Utility: 7)

A collection of guard hooks and skills for Claude Code that block destructive shell commands, prevent secret leakage, and automate common repository chores like commit messages and PR descriptions. It's notable for its rigorous testing methodology, including a public tamper-case corpus, unit tests, and versioned eval transcripts that demonstrate reliability and effectiveness.

Target audience: developers using Claude Code, DevOps engineers, security-conscious teams

Repository: https://github.com/skillkeel/skillkeel-starter · Python · MIT · 2 stars

View on Hacker News