simplify
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of repository source code, diffs, and local instructions which may contain malicious content designed to influence agent behavior.
- Ingestion points: The skill reads code diffs, caller functions, repository-specific instructions, and test documentation as defined in
SKILL.md. - Boundary markers: The instructions employ a structured multi-reviewer synthesis process and an explicit requirement that intended behavior must remain unchanged.
- Capability inventory: In 'Apply mode', the skill has the capability to write edits to the filesystem and execute local commands for testing, linting, and type checking.
- Sanitization: There is no explicit sanitization or filtering of the ingested source code content before processing.
Audit Metadata