agentsclimarketplace

Comfyui workflow guardian

Skill yigityildiz0/universal-ai-skill-library/skills/archive-variants/comfyui-workflow-guardian

Professional ComfyUI workflow auditing, refactoring, runtime stabilization, model-chain analysis, and performance tuning for Flux and other pipelines on.From its SKILL.md

Install
npx -y skills add yigityildiz0/universal-ai-skill-library --skill comfyui-workflow-guardian

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

3 things to look at

  • 25 days oldThe repository was created 25 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

4.4 KB, 882 tokens by cl100k_base, as published. Nobody here has run it

ComfyUI Workflow Guardian

Use this skill to work on ComfyUI like an infrastructure and workflow engineer, not like an improvising prompt tweaker.

Core Rules

  • Back up the target workflow before editing it.
  • Change only the workflow the user explicitly names. Do not sync other workflows unless asked.
  • Inspect runtime failures before editing the workflow. A dead backend is not a workflow bug.
  • Prefer one precise change over many speculative changes.
  • Keep optional modules behind switches so they do not add cost when disabled.
  • Preserve working post-processing sections unless the user explicitly wants them changed.
  • Prefer validated architecture rules over folklore or one-off community claims.

Workflow

1. Classify the task

Classify the request before touching files:

  • Runtime failure: startup crash, queue crash, Failed to fetch, access violation, provider error
  • Structure failure: broken links, missing files, overlapping nodes/groups, unreadable layout
  • Quality failure: identity drift, anatomy defects, artifacts, color shifts, over-smoothing
  • Performance failure: generation too slow, unnecessary modules active, too many second-pass steps
  • Architecture change: adding or comparing LoRAs, identity modules, refiners, control branches

2. Read the right reference file

3. Validate before editing

Run the bundled checker on the target workflow:

python scripts/check_comfyui_workflow.py "C:\path\to\workflow.json" --comfy-root "C:\path\to\ComfyUI"

Use it before and after edits. Treat these as hard failures until explained:

  • missing node references
  • missing model or LoRA files
  • overlapping nodes
  • overlapping groups

4. Apply architecture rules

Follow these stable ordering rules unless source documentation for a specific module contradicts them:

  • Put the base model loader first.
  • Put general LoRAs that shape the base image before the main sampler.
  • Put identity model layers after the general LoRA chain when they modify the final model directly.
  • Put second-pass face-transfer or inpainting systems after the base image is generated.
  • Put detail or sharpening passes after identity transfer, not before.
  • Keep manual size overrides, CFG, and quality toggles in a compact quick-access area, but leave original nodes in place if the workflow already depends on them.

5. Change conservatively

  • Tighten one branch at a time.
  • If a post-pass copies unwanted props, hands, or expressions, reduce what that pass is allowed to transfer instead of raising its strength blindly.
  • If identity is weak, test the identity modules separately before combining them.
  • If performance is poor, reduce second-pass cost before degrading the main generation stage.
  • Do not switch models, precision modes, or providers without evidence that the current setting is the cause.

6. Validate after editing

After edits:

  • run the workflow checker again
  • confirm the target files exist on disk
  • confirm the launcher and runtime still start
  • confirm the active workflow still loads cleanly
  • summarize only the actual changes, default values, and residual risks

Operating Defaults

Use these defaults when the user does not specify otherwise:

  • Keep the user's main workflow as the only edit target.
  • Keep the runtime conservative and stable on Windows.
  • Treat second-pass identity transfer as optional and isolated.
  • Favor clean, non-overlapping layout and short notes over decorative complexity.
  • Favor explicit validation over guesswork.

Output Requirements

When finishing a task with this skill:

  • name the exact files changed
  • name the critical node or runtime settings changed
  • state what was validated
  • state what remains uncertain

What ships with it: 6 files

13.0 KB alongside SKILL.md, 1 of them executable

agents/

scripts/

Keep looking

Skills are one crate of 326,569. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.