restyle
Pass
Audited by Gen Agent Trust Hub on Nov 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user data which is interpolated into ComfyUI workflows for video generation.
- Ingestion points: User-supplied text prompts (pos.text, e_pos.prompt) and video files (src.video, 199.file) in SKILL.md and workflow files.
- Boundary markers: None; user inputs are directly assigned to node parameters in the JSON workflows.
- Capability inventory: The bundled scripts (comfy_client.py) can read local files, perform network POST/GET requests to a user-defined COMFY_URL, and write downloaded files to the local file system.
- Sanitization: No evidence of input validation or escaping for the user-provided text or media metadata.
- [COMMAND_EXECUTION]: The skill executes bundled Python utility scripts (scripts/comfy_client.py and scripts/preflight.py) to interact with a remote or local ComfyUI API. These scripts perform network operations and file management essential for the skill's documented functionality.
- [EXTERNAL_DOWNLOADS]: The references/requirements.md file lists external GitHub repositories for ComfyUI extensions, such as Kosinkadink/ComfyUI-VideoHelperSuite and Fannovel16/comfyui_controlnet_aux. these are provided as documentation for server-side setup and do not involve automated runtime downloads by the skill itself.
Audit Metadata