The Agent Skills Directory

[SAFE]: The skill implements a maker/checker pattern for code review. Its instructions promote skepticism and require manual verification of test results rather than trusting previous claims.
[COMMAND_EXECUTION]: The instructions require the agent to execute project test and lint commands. This is the intended primary function for verifying code integrity and does not involve hardcoded or suspicious commands.
[PROMPT_INJECTION]: The skill processes untrusted input from issue descriptions and implementer summaries. This is an indirect injection surface; however, the skill's explicit instruction to verify evidence and reject by default acts as a robust behavioral constraint against manipulation.

loop-verifier