ComfyUI Workflow Analysis Skill

ComfyUI Workflow Analysis

Understand workflow structure and node connections.

MCP Tools

| Tool | When to use | |------|-------------| | comfy_read | First step - get human-readable summary of any workflow | | comfy_trace | See a specific node's inputs and outputs with slot names | | comfy_upstream | Find all nodes feeding INTO a target node | | comfy_downstream | Find all nodes FED BY a source node |

Workflow

Summarize: comfy_read(workflow) → pattern, key params, flow
Deep dive: comfy_trace(workflow, node_id) → specific connections
Follow data: comfy_upstream/comfy_downstream → trace the graph

What comfy_read Returns

## Pattern: Flux txt2img +LoRA

## Key Parameters
  model: flux-dev.safetensors
  steps: 20
  cfg: 3.5

## Variables (3)
  $model <- Node 1 (CheckpointLoader)

## Flow
  [image] 1 -> 5 -> 12 -> 15

What comfy_trace Returns

[Node 577] WanVideoSampler

  INPUTS:
    [0] model <- Node 608 (GetNode, slot 0)
    [1] image_embeds <- Node 1179 (WanVideoVACEEncode, slot 0)

  OUTPUTS:
    [0] samples -> Node 595 (WanVideoDecode, slot 1)

CLI Equivalents

If MCP isn't available, use CLI commands:

| MCP Tool | CLI Command | |----------|-------------| | comfy_read | python we_vibin.py analyze workflow.json | | comfy_trace | python we_vibin.py trace workflow.json NODE | | comfy_upstream | python we_vibin.py upstream workflow.json NODE | | comfy_downstream | python we_vibin.py downstream workflow.json NODE |

Setup

Requires MCP server. See MCP_SETUP.md.

Agent Skills: ComfyUI Workflow Analysis

Install this agent skill to your local

Skill Files