improve
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is purely instructional and contains no executable code, remote script downloads, or obfuscated content. The instructions emphasize maintaining behavior and seeking user consent for significant changes, which are safe practices.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted code as input, which constitutes a vulnerability surface. 1. Ingestion points: Target code files read during the process step in SKILL.md. 2. Boundary markers: Instructions are provided to ask the user before continuing if the scope changes or touches a public interface. 3. Capability inventory: The skill involves reading and writing files to perform refactoring. 4. Sanitization: No explicit input sanitization or validation of the target code is specified within the skill instructions.
Audit Metadata