refactoring
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted code provided by users, which creates a vulnerability surface where malicious instructions embedded in the input code could be interpreted by the agent during the refactoring process.
- Ingestion points: The agent accepts source code and refactoring goals directly from the user as described in the Usage section of SKILL.md.
- Boundary markers: The skill lacks instructions for the agent to use delimiters or guardrails to isolate the untrusted input code from its own operational instructions.
- Capability inventory: The workflow directs the agent to modify files and execute generated test suites, providing a pathway for injected instructions to result in file system changes or command execution.
- Sanitization: There are no requirements for validating or sanitizing the input code before it is refactored or tested.
- [DYNAMIC_EXECUTION]: The workflow requires the agent to generate and run code dynamically.
- Evidence: Step 5 in SKILL.md instructs the agent to "Execute the existing test suite" and "generate lightweight unit tests covering the refactored paths," which involves the creation and execution of scripts at runtime based on the input code.
Audit Metadata