The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides shell commands using standard utilities (find, grep, sed, comm) to synchronize the skills directory with the README.md documentation.
[SAFE]: The skill includes explicit security and privacy constraints for its 'routine' audit mode, instructing the agent to never print prompt bodies, environment values, or configuration values to the output.
[SAFE]: The skill processes other local files to perform its audit tasks. While this creates a surface for indirect prompt injection, the risk is minimized by the skill's narrow scope and instructions to report only identifiers and field names rather than content.

skill-auditor