The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied text to generate audio prompts, using specific dialogue tags to wrap spoken content. This creates a surface where instructions could potentially be embedded in the data provided to the model.
Ingestion points: User-supplied spoken words are integrated into production prompts as described in SKILL.md.
Boundary markers: The instructions mandate the use of Dialogue: <d>[English] ... </d> tags and [Integrated audio description: ...] brackets to delimit content.
Capability inventory: The skill has the capability to execute a local preparer script (prepare_workflow.py) and perform network operations to communicate with a ComfyUI server instance (e.g., http://127.0.0.1:8188).
Sanitization: The documentation does not specify explicit sanitization or filtering of the user-provided text before interpolation.
[EXTERNAL_DOWNLOADS]: The skill's verification documentation in references/verified-behavior.md includes links to external resources on Reddit, Hugging Face, and the official ComfyUI blog. These links are used to document the provenance and behavior of the audio generation technique and target well-known technology services.
[COMMAND_EXECUTION]: The skill utilizes a shared script named prepare_workflow.py (via a delegated reference to the minimax-h3-comfyui skill) to generate the API JSON files required for ComfyUI execution. This involves local command execution to prepare execution graphs.

minimax-h3-audio