The Agent Skills Directory

[SAFE]: No malicious patterns or security risks were identified. The skill follows best practices for security auditing, emphasizing evidence-based analysis and git history validation.
[COMMAND_EXECUTION]: The skill utilizes Bash to execute standard utilities such as git, grep, find, and wc. These tools are used appropriately within the local environment to analyze codebase diffs and calculate blast radius, with no evidence of command injection or malicious shell execution.
[PROMPT_INJECTION]: This skill serves as a surface for indirect prompt injection because it ingests and processes untrusted data from pull requests and commit diffs. However, the methodology provides a structured adversarial framework for the agent, which mitigates the risk of being influenced by instructions embedded within the code changes being reviewed.

differential-review