The Agent Skills Directory

[SAFE]: The skill is a comprehensive collection of markdown-based instructions and reference materials for conducting code reviews. It establishes clear trust boundaries, specifically instructing agents to treat reviewed content as untrusted data and forbidding them from following instructions found within reviewed diffs.
[SAFE]: All external references and methodologies (e.g., Composer review, security patterns, reliability patterns) are standard engineering practices. The skill includes sophisticated 'Skeptic' and 'Red-Team' protocols to improve review accuracy and prevent false positives.
[SAFE]: There is no executable code (Python/Node.js) included in the skill; it operates entirely through prompt-based instructions and reference routing. It explicitly prohibits secret exposure, hardcoded credentials, and unauthorized network operations in its internal specifications (SPEC.md) and trust boundaries (reviewer-trust-boundary.md).

code-review