The Agent Skills Directory

[SAFE]: The skill instructions and all provided reference files are focused exclusively on performing technical code reviews. The skill detects the programming language in a local repository using git and grep commands, which are standard for development agents. It does not attempt to access sensitive files, communicate with external servers, or bypass safety guidelines.
[INDIRECT_PROMPT_INJECTION]: As a code review tool, the agent is designed to ingest and analyze untrusted code changes (diffs). This inherently presents an attack surface for indirect prompt injection. However, the skill explicitly mandates a 'Review Verification Protocol' that requires the agent to verify findings against the actual source code and cross-reference identifiers throughout the codebase. This protocol acts as a robust defense against adversarial content in the code being reviewed. (Severity: LOW)

do-backend