comfy-diffusion-reference
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill content consists purely of technical documentation and API references for the comfy_diffusion library. It does not contain executable code, obfuscated content, or instructions to bypass safety protocols.\n- [INDIRECT_PROMPT_INJECTION]: The library documented here processes user-supplied text inputs for generation tasks, which represents a functional surface for potential indirect prompt injection.\n
- Ingestion points: Parameters like prompt, text, clip_l_text, tags, and lyrics in various module signatures within SKILL.md.\n
- Boundary markers: The provided API reference does not document specific delimiters or sanitization instructions for these inputs.\n
- Capability inventory: The library offers broad capabilities for model loading, latent manipulation, and image/video/text generation as described in the reference.\n
- Sanitization: Input sanitization and validation are not explicitly detailed within this technical reference documentation.
Audit Metadata