The Agent Skills Directory

[SAFE]: The skill is designed to increase safety by intercepting and blocking potentially dangerous Git commands. It provides a bash script that validates commands against a list of known destructive patterns.
[COMMAND_EXECUTION]: The skill instructs the agent to install a bash script (scripts/block-dangerous-git.sh) and apply executable permissions (chmod +x). This is the intended behavior for deploying the hook and is performed on the skill's own bundled resources.
[PERSISTENCE]: The skill modifies ~/.claude/settings.json or a project-level equivalent to register the PreToolUse hook. While this creates a persistent configuration change, it is the primary purpose of the skill as documented and serves a defensive role.
[DATA_EXPOSURE]: No sensitive data exposure was detected. The script processes tool input locally using jq and grep without making network requests or accessing credentials.

git-guardrails-claude-code