The Agent Skills Directory

[SAFE]: The skill defines a defensive strategy for processing untrusted external code review feedback, requiring technical verification and skepticism before any implementation, which serves as a mitigation against indirect prompt injection.
[COMMAND_EXECUTION]: Employs standard version control commands (git rev-parse, git log) and instructs the agent to run project-specific verification commands (tests, builds) to provide evidence for task completion. These operations are restricted to the local environment and the specific development context.
[PROMPT_INJECTION]: Instructions focus on technical rigor and avoid social performativity, which strengthens the agent's adherence to its primary development tasks without introducing bypass mechanisms.

code-review