The Agent Skills Directory

[SAFE]: The skill functions as an orchestrator (meta-skill) that delegates actual code analysis to a predefined sequence of atomic skills (e.g., review-diff, review-security). It does not contain executable code or perform dangerous system operations.
[SAFE]: Interaction policies are well-defined, requiring user confirmation for the review scope and providing clear defaults to prevent ambiguous execution.
[SAFE]: No evidence of prompt injection, obfuscation, or data exfiltration attempts was found in the provided files. The skill focuses solely on aggregation and deduplication of findings from internal tools.

review-code