The Agent Skills Directory

[SAFE]: The skill provides structured instructions for an AI agent to act as a technical advisor. It emphasizes grounding recommendations in the existing codebase using standard platform tools like Read, Grep, and Glob to ensure accuracy.
[SAFE]: The skill includes robust logic for 'pushing back' on user assumptions and 'stress-testing' proposals, which encourages a critical evaluation of inputs rather than blind obedience to potentially malicious or misguided instructions.
[SAFE]: The skill handles data from the local repository (code and documentation) as part of its primary function. While processing untrusted data is a surface for indirect prompt injection, the skill's requirements to 'verify claims against the code' and 'surface contradictions' provide inherent behavioral safeguards.

advisor