windsurf-refactoring-large
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to coordinate operations that span hundreds or thousands of files, creating a significant surface area for instructions embedded in project data to influence agent behavior.
- Ingestion points: The agent is instructed to perform deep analysis on the entire codebase, mapping all affected files and dependencies (SKILL.md, implementation.md).
- Boundary markers: There are no specific instructions or delimiters provided to ensure the agent ignores or sanitizes instructions embedded in the source code, comments, or documentation it processes.
- Capability inventory: The skill operates with powerful tools including
Bash(cmd:*),Write, andEdit, which could be misused if the agent is misled by content within the files it refactors. - Sanitization: The implementation guide does not include steps for sanitizing or validating external content before it is processed by the AI context.
Audit Metadata