adapt
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill inherits an attack surface for indirect prompt injection by processing untrusted layout/UI data.
- Ingestion points: Untrusted code or files are ingested via the
$ARGUMENTSplaceholder inSKILL.md. - Boundary markers: There are no boundary markers or instructions telling the agent to ignore prompt instructions embedded inside the target UI code.
- Capability inventory: The skill possesses file modification capabilities, instructing the agent to 'edit code directly'.
- Sanitization: No input sanitization or verification mechanism is specified for the incoming UI code arguments.
Audit Metadata