The Agent Skills Directory

[SAFE]: The skill consists entirely of instructional documentation and best practices for configuring AI agents. It does not contain executable code, remote script downloads, or network operations.
[COMMAND_EXECUTION]: The skill mentions standard development commands such as npm test, pytest, and mkdir. These are presented as examples of verification and project bootstrapping steps and do not represent unauthorized or dangerous command execution.
[INDIRECT_PROMPT_INJECTION]: While the skill provides instructions for agent behavior, it does not involve the ingestion or processing of untrusted external data that could lead to injection attacks. The capability tier is restricted to standard file management and testing tasks typical for a coding agent.
[DATA_EXFILTRATION]: There are no patterns suggesting data exfiltration. The instructions focus on local project management using files like tasks/todo.md and tasks/lessons.md for internal tracking.

agent-engineering