The Agent Skills Directory

[SAFE]: The skill's primary purpose is defensive. It provides a vet.sh script that uses standard grep patterns to identify common indicators of compromise in other software (e.g., searching for 'curl', 'sudo', 'base64', or 'API_KEY').
[COMMAND_EXECUTION]: The vet.sh script executes shell commands to perform its auditing task. These commands (grep, basename, etc.) are used as intended for a security scanner and do not pose an inherent risk to the host environment outside of the auditing context.
[PROMPT_INJECTION]: The SKILL.md file contains instructions for the agent to follow a specific vetting process. These instructions are standard operational guidelines and do not attempt to bypass agent safety filters or override core system behavior.
[DATA_EXPOSURE]: While the script searches for patterns related to credentials (e.g., 'API_KEY', 'SECRET'), it does so to flag them in other skills during an audit, rather than exposing the current user's secrets.

skill-vetting