The Agent Skills Directory

[SAFE]: The skill's operations are confined to the local repository. It reads from lib/ and knowledge/ directories and writes documentation and plans to docs/. There are no network calls, external downloads, or attempts to access sensitive system files or credentials.
[SAFE]: The skill manages user input through a structured dialogue protocol to refine project requirements. While it generates 'executable plans' that include shell commands (e.g., grep, lint), these are stored in documentation files for future reference and are not executed by the skill itself, posing no immediate execution risk.

grill