The Agent Skills Directory

[SAFE]: The skill implements a gated workflow for feature development that requires explicit user approval before advancing from planning to implementation. This design incorporates a human-in-the-loop security model.- [SAFE]: File system operations are restricted to a specific directory (plans/) and include instructions for sanitizing user-provided feature titles to remove path-unsafe characters, preventing path traversal or file system manipulation.- [SAFE]: The skill uses internal platform tools ($write-a-prd, $prd-to-plan, $code-review) and does not perform any external network requests or download untrusted code from the internet.

deliver-feature