The Agent Skills Directory

[SAFE]: The primary function of this skill is to provide a reference framework for code reviews based on established engineering practices. It contains no malicious logic or obfuscated scripts.
[COMMAND_EXECUTION]: The instructions direct the agent to use standard git diff commands to collect data for review. This behavior is transparent and restricted to repository analysis tools necessary for the skill's purpose.
[REMOTE_CODE_EXECUTION]: Static analysis flagged the presence of eval and exec keywords. These appear exclusively within the "AI 审查执行指南" (AI Review Execution Guide) section as negative examples (anti-patterns) for the reviewer to detect in target code. They are not instructions for the agent to execute code dynamically.
[SAFE]: The skill has a potential surface for indirect prompt injection as it processes external code (ingestion point: git diff output). However, it does not possess dangerous write or network capabilities that could be exploited. The skill itself serves as a defensive measure by providing a capability inventory for the agent to specifically inspect code for security flaws, including injections and hardcoded secrets.

review-code