The Agent Skills Directory

[SAFE]: The skill provides a structured approach for repository orientation that prioritizes safety and context cleanliness. It explicitly forbids reconnaissance subagents from performing write operations ('recon is read-only') and emphasizes the use of scoped, independent subagents to minimize the risk of accidental execution or data leakage. The use of 'pnpm' and 'grep' mentioned in the documentation are provided as examples of standard development reconnaissance and do not constitute unsafe command execution instructions. No evidence of obfuscation, credential harvesting, or unauthorized remote access was detected.

codebase-exploration