Skillv1.1.0

ClawScan security

Comfyui Workflow · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

BenignFeb 28, 2026, 3:47 AM

Verdict: benign
Confidence: high
Model: gpt-5-mini
Summary: The skill's code, instructions, and requirements are consistent with a local ComfyUI workflow runner — it asks for only local ComfyUI connection info and optional token and does not request unrelated credentials or external installers.
Guidance: This skill appears to be what it says: a local ComfyUI workflow executor. Before installing or running it: 1) Ensure COMFY_HOST points to a ComfyUI server you control or trust (the skill will connect to whatever host you configure). 2) Only provide a ComfyUI token if the server requires authentication; tokens are scoped to that server — do not reuse unrelated secrets. 3) Inspect the included scripts (scripts/comfy_run.py and scripts/comfy_api.py) yourself — they are present in the bundle and implement the network calls; confirm there are no unexpected external hosts in their code. 4) Use --dry-run or --inspect first (as the README suggests) to preview the API payload before executing. 5) Be aware the skill writes/updates MODELS_INVENTORY.md and WORKFLOWS_SUMMARY.md as a local knowledge cache — if you don't want persistent changes, run in a disposable/sandbox directory or back up those files first. 6) If you want extra caution, run the skill in an isolated environment (container or VM) so GPU/model loading and file writes are contained.

Review Dimensions

Purpose & Capability: okThe name/description claim a universal ComfyUI workflow executor; the repo contains a CLI wrapper (comfy_run.py), an API/WebSocket client (comfy_api.py), and many workflow JSONs. Declared requirements (Python 3.10+, websocket-client, requests, a running ComfyUI server, optional token) match the stated purpose and nothing extraneous (no cloud credentials, unrelated binaries, or unusual config paths) is requested.
Instruction Scope: noteSKILL.md confines runtime activity to converting workflow JSON → ComfyUI API payloads, uploading local input files, monitoring execution over a WebSocket to the COMFY_HOST, and downloading outputs to a local output directory. It instructs agents to update local caches (MODELS_INVENTORY.md, WORKFLOWS_SUMMARY.md) with the user's environment — this is expected for a local executor but is persistent file writing and should be noted. There are no instructions to contact external endpoints beyond the configured ComfyUI host.
Install Mechanism: okThere is no install spec; this is instruction+script based. Dependencies are standard Python packages (websocket-client, requests) recommended by SKILL.md. No archive downloads or external install URLs are used by the skill files provided.
Credentials: okThe skill requests no required environment variables or credentials by default. COMFY_HOST and a ComfyUI Token are optional and clearly documented; those are proportionate for connecting to an authenticated ComfyUI server. No unrelated secrets or multiple external-service tokens are asked for.
Persistence & Privilege: notealways:false (normal). The SKILL.md explicitly encourages updating MODELS_INVENTORY.md and WORKFLOWS_SUMMARY.md as a local knowledge cache — this creates persistent files under the skill directory. That persistence is expected for this use case but users should be aware the agent is instructed to write/update local files.