yutianxiao6/comfyui-mcp / skills/comfyui-workflow
comfyui-workflow
VerifiedBuild, inspect, modify, validate, run, and troubleshoot ComfyUI workflows through ComfyUI MCP tools. Use for requests involving ComfyUI nodes, models, workflow JSON, graph connections, queue execution, generated outputs, canvas synchronization, LoRA, ControlNet, inpainting, upscaling, image-to-image, video, audio, 3D, or custom nodes.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 5.8 KB |
| agents/openai.yaml | config | 236 B |
| references/execution-and-files.md | markdown | 2.6 KB |
| references/graph-contract.md | markdown | 2.4 KB |
| references/troubleshooting.md | markdown | 2.7 KB |
| references/workflow-patterns.md | markdown | 2.7 KB |
references/graph-contract.md
View on GitHub# ComfyUI graph contract
Use ComfyUI's API-format workflow object for managed graphs. Do not confuse it with the visual canvas JSON exported by the frontend.
## API-format shape
Each top-level key is a string node ID. Each value contains the exact installed `class_type`, an `inputs` object, and optional `_meta` data:
```json
{
"3": {
"class_type": "ExampleNode",
"inputs": {
"literal_input": 1.0,
"linked_input": ["2", 0]
},
"_meta": {"title": "Readable title"}
}
}
```
A link is a two-item array `[source_node_id, source_output_index]`. The output index is zero-based. Node IDs remain strings even when they contain only digits.
## Schema discovery
Treat `comfyui_get_node_type` as the contract for an installed node:
- inspect required and optional input groups;
- preserve exact input names and enum values;
- inspect output types in order before choosing an output index;
- note whether the class is an output node;
- re-check custom nodes instead of relying on remembered versions.
Use `comfyui_list_node_types` to narrow the candidate set. Search by capability, then inspect candidates individually. Use `comfyui_list_models` for model-backed enum inputs because filenames differ per installation.
## Graph construction rules
- Add provider nodes before consumers so IDs and links are easy to reason about.
- Connect only type-compatible outputs and inputs according to the live schemas.
- Supply every required input either as a literal or as a valid link.
- Keep at least one reachable output node when the workflow is intended to execute and save or preview a result.
- Do not create cycles.
- Keep titles descriptive, but use node IDs—not titles—as tool arguments.
## Revisions
Every managed mutation produces a new integer revision. Pass the last returned revision as `expected_revision` on the next edit. A mismatch means someone or something edited the workflow after it was read.
On conflict:
1. call `comfyui_get_workflow` again;
2. compare the current graph with the intended edit;
3. reapply only the missing change with the current revision.
Do not overwrite a conflicting graph with a stale local copy.
## Import boundary
`comfyui_import_workflow` expects API-format JSON. If the user provides visual workflow JSON, explain the difference and convert it only when enough information is available to produce an equivalent API graph. Validate the imported result live before execution.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |