comfyui-media-workflows

What it does

Turns a creative or technical brief into a production-ready ComfyUI workflow package, treating ComfyUI as a graph-based media production runtime. Produces reproducible, reviewable, and safe-to-run graphs and metadata for image, video, upscale, inpaint, conditioning, and batch generation tasks.

How it works

• Start from deliverable and constraints to choose graph topology, models, nodes, and runtime; do not paste a generic workflow first.
• Maintain two artifacts when possible: workflow.json (UI save format with layout and notes) and workflow_api.json (API-format graph for execution).
• Keep metadata to recreate a run: input files with hashes, model identifiers/hashes, custom-node revisions, ComfyUI version, parameters, seed policy, hardware/runtime, output paths, QA notes.
• Prefer fewer custom nodes and model dependencies due to supply-chain risk.
• Decide whether to embed workflow metadata in outputs and note potential leakage concerns.
• Produce a short production brief first with sections Deliverable, Creative intent, Inputs, Runtime, Model inventory, Custom-node inventory, Rights and safety, Reproducibility target, Handoff target.
• Validate API workflows using the bundled preflight helper before runtime validation, including checks for node existence, input names, and graph integrity.
• For local execution, upload inputs, submit the API-format graph, monitor status, and archive run metadata. For cloud, verify cloud-specific details.
• Maintain a production workflow card with comprehensive metadata (runtime, models, custom nodes, inputs, QA status, known limits).

When to use it

Use when a task involves ComfyUI workflow JSON/API graphs, model and custom-node inventories, reproducibility, provenance, safety/rights review, execution hygiene (local or cloud), or production QA. Begin with the production brief and follow the two-artifact and metadata practices.

What it can touch

• No explicit touch list is provided here beyond the guidance to manage workflow.json and workflow_api.json, and to reference inputs, models, custom-nodes, and runtime details. The skill emphasizes documenting and organizing these artifacts rather than performing runtime operations itself within this description.

Caveats

• Do not assume a specific checkpoint, hosted Comfy service, custom node set, or cloud vendor unless named.
• Re-check model, license, and cloud facts at production time due to frequent updates.
• Do not provide license status, exact approvals, or unknown hashes; mark unknowns plainly in production cards as instructed.