The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires the execution of shell commands to manage the code review lifecycle. This includes using git rev-parse and git log to identify relevant code changes and the execution of project-specific build and test commands to satisfy 'Verification Gates' before any status claims can be made.
[INDIRECT_PROMPT_INJECTION]: The skill establishes a protocol for handling code review feedback from untrusted external sources. While this presents an injection surface, the skill includes explicit mitigation instructions requiring technical verification against the codebase, regression testing, and reasoned pushback against incorrect or unnecessary suggestions (YAGNI), which serves as a robust security control.
[PROMPT_INJECTION]: The instructions mandate a specific technical persona by explicitly forbidding standard AI conversational norms such as 'performative agreement' and expressions of gratitude (e.g., 'Thanks', 'You're right'). While these instructions override default behavioral patterns to prioritize technical rigor, they are procedural constraints rather than malicious attempts to bypass safety filters or exfiltrate data.

code-review