refactoring-planner
Pass
Audited by Gen Agent Trust Hub on Aug 15, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill uses the
[PASTE CODE]variable to include untrusted user code in the prompt template withinSKILL.md.\n - Ingestion points: User-provided code is interpolated into the prompt text used by the agent.\n
- Boundary markers: Input is enclosed in
<code>tags, which provide weak isolation and can be bypassed by adversarial content.\n - Capability inventory: The agent utilizing this skill likely has access to standard tools such as the shell and filesystem.\n
- Sanitization: No validation or sanitization is performed on the user-supplied code to mitigate the risk of embedded malicious instructions.
Audit Metadata