refactor
Pass
Audited by Gen Agent Trust Hub on Feb 21, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- PROMPT_INJECTION (LOW): The skill uses directive instructions such as 'Invoke IMMEDIATELY' and 'Do NOT explore first' to bypass the agent's standard exploration logic. It also presents an indirect prompt injection surface by processing external code. [1] Ingestion points: User-provided source files in the codebase. [2] Boundary markers: The workflow is constrained by predefined smell categories and principles. [3] Capability inventory: The skill executes local Python scripts to orchestrate analysis agents. [4] Sanitization: No explicit sanitization of code content is documented.
- COMMAND_EXECUTION (SAFE): The skill invokes the local Python module 'skills.refactor.refactor' to perform its core orchestration logic, which is standard behavior for script-based skills.
Audit Metadata