chat-widget
Pass
Audited by Gen Agent Trust Hub on Aug 10, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill architecture establishes a surface for indirect prompt injection via user-generated chat messages that are processed and displayed to admin users.
- Ingestion points: User messages are ingested through the support_messages table via WebSocket ChatChannel and REST API endpoints described in SKILL.md.
- Boundary markers: The instructions do not define boundary markers or delimiters to isolate user-supplied message content from other instructions.
- Capability inventory: The system includes capabilities for database persistence, real-time WebSocket broadcasting, and triggering background email notification jobs.
- Sanitization: Sanitization logic is missing from the provided pseudocode and backend implementation examples (Ruby, PHP), although the React frontend example utilizes implicit framework-level escaping.
Audit Metadata