The Agent Skills Directory

[SAFE]: The skill performs a static analysis of a repository's internal documentation (such as CLAUDE.md and README.md) and codebase to assess methodology alignment. This is a legitimate diagnostic use case.
[DATA_EXPOSURE]: The skill is designed to read local files to perform its audit functions. No evidence of unauthorized data access, hardcoded credentials, or network exfiltration to external domains was found.
[COMMAND_EXECUTION]: The skill's process includes safeguards, specifically instructing parallel assessor agents to be 'read-only' and to 'modify nothing.' The only proposed write operations involve capturing reports to files like KANBAN_BOARD.md or IDEAS.md, which are standard project management tasks.
[PROMPT_INJECTION]: There are no patterns suggesting attempts to bypass safety filters or override system instructions. The instructions provided are focused on task delegation and synthesis of audit findings.

promode-audit