← Back to Skills Marketplace
image
by
TobeyRebecca
· GitHub ↗
· v1.0.0
· MIT-0
95
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install toby-image
Description
Run local ComfyUI workflows via the HTTP API. Use when the user asks to run ComfyUI, execute a workflow by file path/name, or supply raw API-format JSON; sup...
Usage Guidance
This skill appears to do what it says: edit workflow JSON and queue runs against a local ComfyUI server, plus download model weights into ~/ComfyUI/models/. Before installing or running it: 1) Review the two scripts (comfyui_run.py and download_weights.py) yourself — they will make network requests (to localhost and to URLs you provide) and write files to your home directory. 2) Only provide model URLs you trust (downloading arbitrary URLs can place malicious or large files on disk). 3) Be aware the download script will attempt to install a pget binary to ~/.local/bin if missing. 4) Ensure you are comfortable running commands that clone/install ComfyUI and running the ComfyUI server locally. 5) Note minor metadata inconsistencies (declared required binaries vs registry metadata and a name mismatch); these are bookkeeping issues but review SKILL.md before use. If you want stronger assurance, run the scripts in a controlled environment (container or VM) first.
Capability Analysis
Type: OpenClaw Skill
Name: toby-image
Version: 1.0.0
The skill bundle is classified as suspicious primarily because `scripts/download_weights.py` automatically downloads and executes a binary (`pget`) from a remote GitHub URL (replicate/pget) if it is not found in the local PATH, which constitutes a supply-chain risk. Additionally, `SKILL.md` instructs the AI agent to perform high-risk system operations including `git clone` and `pip install` for environment setup. Finally, `assets/tmp-workflow.json` contains a NSFW image generation prompt, which is unexpected for a standard utility bundle.
Capability Assessment
Purpose & Capability
The skill's name/description (run local ComfyUI workflows) matches the included artifacts: a run script that posts workflow JSON to a local ComfyUI HTTP API and a download script that saves model weights into ~/ComfyUI/models/. The ability to inspect and edit workflow JSON and to download model weights is expected for this purpose. Minor inconsistencies: registry metadata lists no required binaries while SKILL.md metadata notes python3 is required, and the top-level skill name ('image') differs from SKILL.md's declared name ('ComfyUI'). These are metadata mismatches but do not indicate malicious behavior.
Instruction Scope
SKILL.md explicitly instructs the agent to read and modify workflow JSON (prompt/style/seed fields), write a temp workflow file, and run the bundled run script which posts to the local API (127.0.0.1:8188). It also instructs how to install/run ComfyUI if not present and how to run the download script for model URLs. All referenced files/paths (~/ComfyUI, skills/comfyui/assets/tmp-workflow.json, ~/.local/bin for pget) are coherent with the stated goal. Note: editing arbitrary workflow JSON is powerful — workflows may contain user-provided or sensitive prompts; the agent must only modify fields necessary for generation.
Install Mechanism
There is no formal install spec (instruction-only), but code files are bundled and intended to be run locally. The download script may fetch the pget binary from GitHub Releases (https://github.com/replicate/pget/releases/latest/download/<asset>) and will write it to ~/.local/bin if missing; this is a legitimate helper but does write an executable to the user's home directory. Downloading model URLs (arbitrary URLs provided by the user) will write files under ~/ComfyUI/models/<subfolder> and can store large binaries. These behaviors are expected for the functionality but are higher-risk than pure instruction-only skills because they create files and executables on disk.
Credentials
The skill does not request environment variables, credentials, or access to unrelated services. It uses local filesystem paths (home dir, ~/ComfyUI) and network access to the local ComfyUI server and to user-supplied model URLs — all relevant to its purpose. One minor mismatch: SKILL.md metadata signals a dependency on python3 (a reasonable runtime requirement) but the registry requirements block lists none.
Persistence & Privilege
The skill does not request always:true and does not modify other skills or global agent settings. It does create or write to user-local paths (~/.local/bin for pget and ~/ComfyUI/models/ for weights) and will save edited workflows under the skill assets/tmp-workflow.json; these are scoped to the user's home and the ComfyUI install and are proportional to the task. The automatic attempt to install pget into ~/.local/bin is a persistent change the user should be aware of.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install toby-image - After installation, invoke the skill by name or use
/toby-image - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
- Initial release of the toby-image (ComfyUI) skill.
- Enables running local ComfyUI workflows via the HTTP API, including user-supplied workflow files or raw JSON.
- Supports editing workflow prompts, styles, and seeds before execution.
- Allows downloading model weights from user-provided URLs directly into ComfyUI's model folders.
- Sends generated images back to the user after every successful run.
Metadata
Frequently Asked Questions
What is image?
Run local ComfyUI workflows via the HTTP API. Use when the user asks to run ComfyUI, execute a workflow by file path/name, or supply raw API-format JSON; sup... It is an AI Agent Skill for Claude Code / OpenClaw, with 95 downloads so far.
How do I install image?
Run "/install toby-image" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is image free?
Yes, image is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does image support?
image is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created image?
It is built and maintained by TobeyRebecca (@tobeyrebecca); the current version is v1.0.0.
More Skills