comfyui
Pass
Audited by Gen Agent Trust Hub on Nov 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill describes an attack surface for indirect prompt injection via the processing of community-sourced ComfyUI workflows.
- Ingestion points: The agent ingests external data through the
workflow_jsonandworkflow_pathparameters as instructions inSKILL.md. - Boundary markers: There are no explicit prompt delimiters or boundary markers defined in the skill for isolating external workflow content.
- Capability inventory: The skill utilizes
comfyui_image,comfyui_video, andcomfyui_musictools to execute logic on a ComfyUI server. - Sanitization: The skill provides defensive instructions, directing the agent to inspect loader nodes and validate the API format of the JSON before proceeding with execution.
- [EXTERNAL_DOWNLOADS]: The skill mentions that tool responses for missing models may contain download URLs. This is utilized for routine model management and setup within the ComfyUI ecosystem and is disclosed as a standard part of the failure handling protocol.
Audit Metadata