The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands like git rev-parse, git ls-files, fd, and pwd to identify target files and ensure operations are confined to the current git repository.
[COMMAND_EXECUTION]: It may invoke local repository automation tools, specifically just recipes for Markdown formatting and skill metadata verification, with explicit instructions to inspect the commands before execution.
[SAFE]: Implements mandatory guardrail checks that abort execution if the agent is at the filesystem root, the user's home directory, or sensitive agent configuration paths (e.g., ~/.agents or ~/.claude).
[SAFE]: Instructions for installing missing dependencies point to the official vendor's repository (PaulRBerg/agent-skills) using the standard npx package runner.

agent-context-polish