What problem does it solve?
Setting up a ComfyUI model family requires manually hunting down custom nodes, model weights, and a working workflow, which is error-prone and hard to reproduce across machines.
Core Features & Use Cases
- Manifest-driven installation: Run
apply_manifest against a pack's manifest.yaml to idempotently install custom nodes and download model weights, skipping anything already present.
- Pack authoring: Derive a new pack from an upstream installer script or a workflow JSON by extracting model filenames, resolving them to verified Hugging Face URLs, and generating
.bat/.sh installers via npm run packs:gen.
- Validation pipeline: Validate packs with
packs:validate (schema), packs:check-urls (live URL checks), and packs:test (offline idempotency).
- Use Case: A user wants to run a WAN video workflow on a fresh ComfyUI install; the skill applies the pack manifest to fetch all nodes and weights, then points them to the pack's
workflow.json and VRAM notes.
Quick Start
Ask the AI to install the installer pack for a given model family by applying its manifest.yaml with the apply_manifest tool.