The Agent Skills Directory

[SAFE]: The skill implements a verification layer for project artifacts, ensuring that other components of the workflow have produced the required documentation and that human review gates are respected. It operates locally on project files and lacks capabilities for network communication or privilege escalation.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from workspace files (markdown reviews and JSON reports) created by other agents. However, it applies strict, non-executable logic such as counting specific patterns (e.g., checkmarks) and comparing file timestamps. The lack of dangerous tools or command execution capabilities mitigates the impact of potential injection via processed data.

completeness-auditor