The Agent Skills Directory

[SAFE]: The skill is a comprehensive framework for managing agent lifecycle hooks. It provides templates for security, including scrubbing secrets and blocking destructive commands, as well as operational integrity tools for verifying code changes and task commitments.
[COMMAND_EXECUTION]: The provided Python scripts use subprocess.run for legitimate administrative tasks such as checking the agent's credit balance via an internal API and running self-diagnostic tests to ensure hooks are functional.
[EXTERNAL_DOWNLOADS]: The runtime_footer.py script makes HTTP requests to an internal credit API (starchild-credit-api.internal) to display the user's remaining balance. This is a documented feature of the agent's platform and uses internal infrastructure.
[SAFE]: Automated scanner alerts regarding remote code execution and destructive system commands are identified as false positives. These patterns, such as 'curl | bash' and 'rm -rf /', appear exclusively within templates/security_guard_selftest.py as test data. These cases are used to verify that the 'security guard' hook correctly identifies and handles such commands according to its defined policy.

agent-hooks