The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the setup of a local Bash script (block-dangerous-git.sh) that acts as a validator for incoming commands within the Claude Code environment. This is a standard use of the platform's hook system for implementing security guardrails.
[PERSISTENCE]: The skill instructs the agent to modify the project or global settings file (settings.json) to register a PreToolUse hook. While this ensures the safety logic persists across sessions, it is the primary intended function of the skill and aligns with the platform's architectural design for user-defined guardrails.
[PRIVILEGE_ESCALATION]: The setup instructions include a chmod +x command to make the bundled safety script executable. Since this operation is performed on the skill's own distributed script to enable its core functionality, it is considered a safe and expected administrative task.

git-guardrails-claude-code