The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted code diffs, which acts as an ingestion point for external data. This creates a surface for indirect prompt injection where malicious content embedded in the code (e.g., in comments or string literals) could attempt to influence the agent's cleanup logic. The skill lacks specific boundary markers or sanitization steps to isolate this untrusted content. However, the risk is mitigated as the skill does not grant the agent capabilities for network access, file system writes, or subprocess execution.
[SAFE]: The skill instructions are benign and focused on standard code review practices. No patterns related to obfuscation, credential harvesting, or unauthorized privilege escalation were detected.

slop-check