The Agent Skills Directory

[SAFE]: The skill performs only local file system operations, specifically reading from and appending to docs/BACKLOG.md within the project directory.
[SAFE]: No network operations, external data exfiltration, or remote code execution patterns were detected.
[SAFE]: The instructions include a 'public-repo rule' which explicitly directs the agent to generalize sensitive information (like revenue specifics or names) before writing to the backlog, minimizing the risk of accidental data exposure.
[SAFE]: The skill is strictly limited to capturing information ('capture, not analysis') and explicitly forbids starting new research threads or complex tasks, which reduces the potential for autonomy abuse or unintended side effects.

queue