The Agent Skills Directory

[SAFE]: The skill functions as a prompt template to guide the agent through a multi-stage code review process. It defines specific roles and output formats without introducing technical vulnerabilities.
[COMMAND_EXECUTION]: The skill refers to processing git diff and file paths provided in $ARGUMENTS. It expects the agent to use its native environment tools to access this information and does not include any arbitrary or hidden shell commands.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze external data (code and diffs), which is an inherent surface for indirect prompt injection. However, the instruction set includes a specific 'monitoring' stage (the 'Embroidered Uniform Guard') intended to identify misinterpretations, omissions, or errors in the previous review stages, which serves as a logical mitigation against such attacks.

review-sslb