Skip to content
Automation
Command

/gallery

Browse and inspect generated ComfyUI outputs

From plugin
comfyui-mcp
52211 skills4 agents11 commands
Install
$ npx -y skills add artokun/comfyui-mcp --agent claude-code

How it fires

How this command gets triggered: by you, by Claude, or both.

  • Fires itselfClaude auto-loads it when your prompt matches the work.
  • You can call itInvoke it directly when you want it.
  • Slash command/gallery

Context preview

What this command does when you run it.

Browse and inspect generated ComfyUI outputs

Command definition

gallery.md
description: Browse and inspect generated ComfyUI outputs
argument-hint: "Optional filter: 'last N', 'today', 'yesterday', or filename pattern"

/comfy-gallery — Browse Generated Outputs

The user wants to browse images that ComfyUI has generated, inspect their metadata, and optionally re-run or modify workflows.

Instructions

1. **Determine the output directory.** Call `get_system_stats` to check for a custom `--output-directory` argument. If not set, use the default ComfyUI output directory (typically `~/Documents/ComfyUI/output/` or the `output/` folder relative to the ComfyUI installation).

2. **List image files.** Use the Glob tool to find image files in the output directory:

  • Pattern: `**/*.png` (ComfyUI saves as PNG by default)
  • Also check for `**/*.jpg`, `**/*.webp` if needed
  • Sort by modification time, newest first

3. **Apply filters.** The argument is: $ARGUMENTS

Parse the filter:

  • **`last N`** or **`N`**: show only the N most recent images
  • **`today`**: images created today (compare file dates)
  • **`yesterday`**: images from yesterday
  • **Filename pattern**: glob match against filenames (e.g., `ComfyUI_00123*`)
  • **No filter**: default to showing the last 10 images

4. **Present image details.** For each image, show:

  • Filename
  • File size (human-readable: KB/MB)
  • Creation/modification date and time
  • Use the Read tool to display the image thumbnail if the file is a supported image format

5. **Extract workflow metadata.** When the user selects an image or asks for details:

  • Use `get_workflow (action:"from_image")` to extract the embedded workflow JSON from the PNG metadata
  • Show key parameters: checkpoint used, prompt, negative prompt, seed, steps, CFG, sampler, dimensions
  • Offer to visualize the workflow with `visualize_workflow`

6. **Offer actions.** For any selected image, offer to:

  • **Re-run**: execute the same workflow again (new seed for variation)
  • **Modify and re-run**: let the user change parameters (prompt, seed, steps, etc.) and run a modified version
  • **Upscale**: create an upscale workflow using the image as input

Example

User: `/comfy-gallery last 5`

Steps:

  • Find the output directory
  • List all PNG files, sorted by newest
  • Show the 5 most recent with filename, size, and date
  • Display thumbnails
  • Ask if the user wants to inspect any image's workflow

Notes

  • ComfyUI embeds workflow metadata in PNG files by default — this is what `get_workflow (action:"from_image")` reads
  • JPEG and WebP files may not have embedded metadata depending on ComfyUI settings
  • The output directory may contain subdirectories for different workflows or dates
  • If the output directory is empty, tell the user no images have been generated yet
  • Large galleries (100+ images) should be filtered to avoid overwhelming output
Read more
Ships withcomfyui-mcp

The local-first, agent-native control plane for ComfyUI — an MCP server + live sidebar agent that generates images, video and audio, authors and runs workflows, manages models and custom nodes, and edits your live ComfyUI graph in natural language.

Get the whole plugin, auto-invoked
Stats
522
Stars
0
Views
84
Forks
Active
Maintenance
TypeScript
Language
MIT
License
2h ago
Last commit
5mo ago
Created

Repo: artokun/comfyui-mcp