eval-integration
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill serves as a checklist and instruction set for an AI agent to perform code reviews and integration testing. It specifically looks for security best practices such as ensuring no secrets are committed to code and that database schemas follow data integrity rules (e.g., timestamps and cascade deletes).
- [CREDENTIALS_UNSAFE]: The skill explicitly instructs the agent to check for hardcoded secrets in source files and verify the existence of
.env.examplefiles, which aligns with security best practices for credential management. - [DATA_EXPOSURE]: The evaluation passes include checking for 'Secrets in code' to prevent exposure. The instructions do not contain any hardcoded credentials or sensitive data.
- [INDIRECT_PROMPT_INJECTION]: While the skill processes external data (code, schemas, and plans), it does so in a review capacity. It lacks high-risk capabilities like arbitrary command execution or network exfiltration that would make it vulnerable to exploitation via indirect injection. The severity is low/safe.
Audit Metadata