The Agent Skills Directory

Independent Code Assessment: The skill implements a security-conscious workflow by requiring the agent to analyze the code diff and surrounding context before reading potentially biased or malicious PR descriptions, which helps mitigate indirect prompt injection risks by ensuring the agent's initial judgment is anchored in the code itself.
API Surface Verification: It uses a structured procedure to verify that any new public API surface has received formal approval, leveraging tools to inspect issue labels and comments within the repository's ecosystem to ensure compliance with project governance.
Local Capability Invocation: The skill invokes area-specific review agents located within the repository's directory structure. This modular design allows for task specialization while keeping execution scoped to local, project-defined resources.
Orchestrated Multi-Model Review: It utilizes the platform's sub-task capabilities to cross-verify findings across different model families. This enhances the reliability of the code review by aggregating diverse perspectives while remaining within the authorized execution environment.

code-review