distill
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided UI code and content, creating a potential surface for indirect instructions.
- Ingestion points: External UI content and code snippets are ingested via the
$ARGUMENTSvariable defined inSKILL.md. - Boundary markers: The instructions lack specific delimiters or guardrails to ensure the agent ignores instructions potentially embedded within the UI code being refactored.
- Capability inventory: The skill is explicitly permitted to "Edit code directly," providing a write capability based on the processed input.
- Sanitization: No sanitization, escaping, or validation logic is defined for the ingested UI content before processing.
Audit Metadata