comfyui
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes community ComfyUI workflows and recipes, creating an attack surface for instructions embedded in data.
- Ingestion points: The
workflow_json,workflow_path, and community workflow imports inSKILL.mdaccept external data. - Boundary markers: The instructions lack explicit delimitation or warnings to ignore instructions within imported workflows.
- Capability inventory: The skill performs network operations via
POST /promptandGET /history, and writes models to the local file system. - Sanitization: No sanitization process for the incoming JSON workflow structure is described.
- [EXTERNAL_DOWNLOADS]: Recommends downloading resources from community-provided links.
- Evidence: Advises fetching models from 'workflow author's model links' and installing custom nodes from 'author's documented install path'.
- [DATA_EXFILTRATION]: Documents Partner Nodes such as Gemini, Seedance, and MiniMax that require network access and call hosted APIs, which involves transmitting data to external service providers.
Audit Metadata