refine-code
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted repository content, creating an indirect prompt injection surface where malicious code comments or documentation could influence agent behavior.
- Ingestion points: Source files, Git history, ADRs, and repository-specific instructions are read into the agent context (SKILL.md, ARCHITECTURE.md).
- Boundary markers: The skill uses explicit instructions to limit context to user-specified files and relevant neighboring code (SKILL.md, ARCHITECTURE.md).
- Capability inventory: The skill is authorized to modify or delete files and execute shell commands for testing, linting, and building (SKILL.md, ENTROPY.md).
- Sanitization: No specific content sanitization is described for incoming file data.
- [DYNAMIC_EXECUTION]: The skill is instructed to discover and execute repository-native verification tools, compilers, linters, and test suites (ENTROPY.md). This involves executing code or scripts defined within the analyzed repository as part of the verification process.
Audit Metadata