The Agent Skills Directory

[SAFE]: The skill functions as a static analysis tool for AI agent configurations. It validates local agent definitions and delegation prompts without external dependencies, network activity, or remote code execution.
[COMMAND_EXECUTION]: Employs the Bash tool to perform local file discovery and pattern matching using standard utilities like grep and ls. These operations are limited to the .claude directory and are used solely for structural validation.
[PROMPT_INJECTION]: Includes specialized logic to detect and report identity re-definition patterns (e.g., 'You are a...') in other skills. This behavior is a defensive feature that helps prevent unintended agent behavior and ensures role boundaries are respected.

delegation-check