comfyui-workflows
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill demonstrates secure design by explicitly prohibiting the agent from requesting user credentials and instead utilizing established platform authentication. It also includes protective guidelines regarding resource management and budget transparency to prevent accidental overcharging or data misuse.
- [INDIRECT_PROMPT_INJECTION]: The skill defines an attack surface for indirect prompt injection by processing external workflow schemas and user overrides.
- Ingestion points: Workflow graph data and input parameters retrieved via MCP tool outputs as described in SKILL.md.
- Boundary markers: Instructions mandate mapping inputs to specific node IDs and real names, reducing the risk of the agent misinterpreting data as commands.
- Capability inventory: Tools for deployment inspection, GPU management, and inference request submission.
- Sanitization: The instructions enforce the use of schema-valid overrides and adherence to the deployment's exposed input contract.
Audit Metadata