The Agent Skills Directory

[SAFE]: The skill incorporates a robust security scanning process that inspects task documents for high-risk keywords associated with file deletion (rm, unlink), system command execution (sudo, chmod), and data loss (DROP, TRUNCATE) before starting implementation.
[SAFE]: It implements a least-privilege approach for background sub-agents by defining a restricted set of allowed tools in the sub-agent configuration, effectively preventing unauthorized or dangerous actions.
[SAFE]: A mandatory permission check and user approval workflow is required before launching parallel tasks, ensuring human oversight of automated operations.
[SAFE]: The skill enforces the use of external security and best-practice auditors (such as Supabase and Vercel best practice skills) to validate code quality and security standards before any code is committed.

implement