super-code
Pass
Audited by Gen Agent Trust Hub on Sep 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill incorporates a mandatory 'Guardrail check' requiring the agent to verify that correctness and security have not been compromised during code optimization.
- [SAFE]: Language-specific reference files promote modern and secure coding practices, such as using bounds-checked string functions in C and smart pointers in C++, directly reducing the risk of common vulnerabilities.
- [INDIRECT_PROMPT_INJECTION]: The skill operates on untrusted user-provided source code and possesses file-write capabilities. It instructs the agent to suppress explanatory prose, which reduces user oversight during code modifications. However, the requirement for targeted diffs ensures that changes remain reviewable.
- Ingestion points: The skill refactors user-provided source code during development tasks.
- Boundary markers: No explicit delimiters or ignore-markers are defined in the instructions for user data.
- Capability inventory: The agent has file-system write access through targeted patches and standard IDE tool execution.
- Sanitization: The skill requires a mandatory internal check to prevent security regressions during the refactoring process.
Audit Metadata