automateyournetwork/netclaw/comfyui-topology-viz

Claude Code and agent skill by automateyournetwork/netclaw

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Turn a network topology into one stylized, AI-generated still image via a self-hosted ComfyUI instance — reuses the same topology model as threejs-network-viz (any of 8 topology-source integrations, or a freeform description). Use when the operator asks for a stylized, flashy, or AI-generated image/picture/illustration of a network topology. Stills only — no video/animation.

Install

npx vskill@latest install automateyournetwork/netclaw/comfyui-topology-viz

SKILL.md

The same SKILL.md was published earlier as automateyournetwork/netclaw/comfyui-topology-viz.

ComfyUI Network Topology Visualization Skill

Version: 1.0.0 Feature: 120-comfyui-topology-viz Status: Active

Overview

Turns a network topology into one stylized, AI-generated still image via a self-hosted ComfyUI instance — a different rendering path from NetClaw's existing three.js (threejs-network-viz), Blender (blender-3d-viz), and UE5 (ue5-network-viz) skills, which all produce navigable 3D scenes. This skill produces one flat, "flashy" illustration per request instead, reusing the same canonical topology model spec 046's three.js skill already assembles from any of NetClaw's existing topology-source integrations, or a freeform description.

v1 scope is stills only. Video (traffic flybys, packet-tracing animations) and stylized test-result cards are explicitly out of scope for this version — likely follow-on specs, not built here (spec.md FR-016).

Prerequisites

Required
  • A separately running ComfyUI instance — this skill does not install or manage ComfyUI

itself, only connects to one you already have running. Get ComfyUI from https://github.com/comfyanonymous/ComfyUI or the ComfyUI Desktop app.

  • `COMFYUI_URL` in .env (see .env.example) — your ComfyUI instance's endpoint, e.g.

http://127.0.0.1:8000. Required; there is no assumed default (FR-005), because comfyui-mcp's own built-in defaults (8000 for ComfyUI Desktop, 8188 for a manual install) are easy to mix up with your actual instance's port.

  • At least one image-generation checkpoint installed in ComfyUI (Stable Diffusion 1.5, SDXL,

or Flux) — see "If you get 'no usable model found'" below if you haven't installed one yet.

  • The vendored comfyui-mcp server (mcp-servers/comfyui-mcp/), cloned and built via

npm install && npm run build, registered as comfyui-mcp in config/openclaw.json.

Topology sources

Any of NetClaw's existing topology-of-record or lab-emulation integrations — Cisco Modeling Labs, GNS3, containerlab, EVE-NG, Nautobot, NetBox/Infrahub, IP Fabric, or Forward Networks — or a freeform plain-language description requiring no live source at all.

Natural Language Commands

Render a live topology as a stylized still image (User Story 1)
"Give me a stylized AI image of the CML lab topology"
"Render my GNS3 project as a ComfyUI image"
"Make a flashy image of this network"

NetClaw retrieves the topology from the named (or clarified) source, checks what image-generation checkpoints ComfyUI has installed, generates one image, and tells you where it was saved: workspace/output/comfyui-topology-viz/comfyui-<timestamp>-<request-id>.png (plus a sidecar .json recording the prompt and checkpoint used) — and which checkpoint was used (FR-006a).

Sketch a topology without a live source (User Story 3)
"Make a flashy image of this topology: a router r1 connected to a switch sw1"

Same generation pipeline as a live-sourced request, just parsed from your plain-language description instead.

If something isn't available (User Story 2)

Every failure NetClaw can detect gets a specific, distinct message — never a hang or a generic error:

ConditionWhat NetClaw tells you
ComfyUI unreachable at the configured COMFYUI_URL (or comfyui-mcp silently connected to a different instance instead — see Known Limitations)The configured endpoint could not be reached
ComfyUI reachable, but no installed checkpoint is suitable for image generationWhat kind of model to install (SD1.5/SDXL/Flux) and where
ComfyUI itself reports the generation job failedThe generation failed — as distinct from a reachability or missing-model problem
A generation is already runningWait for it to finish (or fail) and ask again — this skill runs one job at a time, never two concurrently
A named topology source is unreachableThe sourcing failure, distinct from anything ComfyUI-side
The topology has zero devicesNothing to visualize — no generation is attempted

There is no NetClaw-imposed timeout on generation itself — real GPU image-generation time varies with model, workflow, and hardware, so a submitted job is tracked to ComfyUI's own completed/failed status rather than being given up on early.

If you get "no usable model found"

Expected on a fresh ComfyUI install with no checkpoints downloaded yet — not a NetClaw bug. Install at least one Stable Diffusion 1.5, SDXL, or Flux checkpoint into ComfyUI's models/checkpoints directory (ComfyUI Manager, or a manual download into that folder), then ask again.

Architecture

ModuleResponsibility
topology_model.pyCanonical Device/Interface/Link/TopologySnapshot types (ported from threejs-network-viz, trimmed of 3D-only concepts)
materials.pyHostname-based device-role inference (ported, trimmed of color tables)
sources.pyOne adapter per topology source (8 live integrations + freeform), plus source-selection disambiguation (ported)
generation_model.pyNEW entities: GenerationRequest, ModelAvailabilityCheck, GeneratedImage, the six-kind GenerationFailure taxonomy
comfyui_client.pyMCP stdio client wrapper around the vendored comfyui-mcp server — discovery, template selection, async submission, no-timeout polling, plus the ControlNet workflow builder and image upload
prompt_builder.pyTopologySnapshot → a bounded-length, role/count-summarized generation prompt
topology_renderer.pyDeterministic (NOT AI) box/line structure diagram, fed to ComfyUI's Canny node as ControlNet conditioning — the structural-accuracy pipeline (research.md §10)
label_overlay.pyBurns real, correct hostname labels onto the completed generation deterministically — Canny-conditioned text is too lossy for Flux to reproduce reliably (research.md §10)
generation.pyOrchestrates the full call sequence and every failure classification; picks the structural ControlNet path when available, falls back to plain txt2img otherwise; enforces the single-in-flight-job guard — unmodified since spec 120 (spec 121 FR-012)
output.pyOverlays labels (if structural path used) and copies the completed image into workspace/output/comfyui-topology-viz/ with a timestamped name + sidecar JSON, never overwriting — unmodified since spec 120
federated_generation.pyNEW (spec 121) — the actual entry point __init__.py now calls. Routes each request between the federated path (below) and generation.run_generation()'s existing pipeline (called as-is, never modified)

See contracts/comfyui-generation-contract.md for spec 120's exact fallback call sequence, and specs/121-federated-topology-viz/contracts/ for the two new federated-stage tool contracts.

The federated path (spec 121)

Same entry point, no new command — every "give me a stylized image" request goes through this skill exactly as before; which path actually produced the delivered image is now visible in the response as generation_path:

generation_pathMeaning
federatedBoth stages ran on an operator-configured federation member (e.g. <your-risk>/viz): a deterministic, correct-by-construction diagram (real role icons, real labels, real connections — no diffusion model involved), then a diffusion image-edit pass that restyled it without altering structure. This is the strongest correctness guarantee this skill can offer (spec 121 SC-001).
federated_partialThe structural diagram (correct, unstyled) was produced on the configured federation member, but the styling stage failed or that half of the member was unreachable — you still get the correct diagram, not nothing, with reason telling you styling didn't complete.
fallbackSpec 120's original Flux+ControlNet+Canny pipeline (unchanged) — used for a freeform request (no real device data for the structural stage to work from) or when the configured federation member itself is unreachable. reason says which.

Both new stages run as separate MCP servers (mcp-servers/topology-diagram-mcp/, mcp-servers/image-style-mcp/) invoked from Border via n2n/tools/call on the live your configured federation member — Border never renders or diffuses anything itself for this path (FR-005). See specs/121-federated-topology-viz/research.md for the full design, including four real gaps found and fixed in the shared federation infrastructure itself (R10) to make this actually work — this was the first working internal n2n/tools/call in the codebase.

If your viz federation member is down: systemctl --user start netclaw-member-<your-risk>-viz.service (service name derived from your own risk/member name, per your N2N federation setup) (it's enabled, so a host reboot brings it back automatically).

Two generation paths (spec 120's fallback pipeline, used when the federated path isn't)

  • Structural (preferred, when Flux + a ControlNet are installed): the topology is rendered as

a plain geometric box/line diagram, ComfyUI's Canny node extracts edges from it, and Flux paints over those edges — the generated image's structure (which device connects to which) is guaranteed accurate because it comes from deterministic code, not the diffusion model. Real hostname labels are overlaid afterward, also deterministically. Verified end-to-end (2026-08-28): exactly the right devices, exactly the right connections, correct legible labels.

  • Plain txt2img (fallback): used only when the ControlNet pipeline's models aren't all

installed. Generates from a text-only prompt with no structural guarantee — the diffusion model is free-associating from a description, not reproducing an accurate diagram. Verified working (produces a real image) but visually confirmed by the user to not resemble the actual topology.

Environment Variables

VariableRequiredPurpose
COMFYUI_URLYesEndpoint of your own already-running ComfyUI instance

Known Limitations

This is the start of the file. Read the full SKILL.md on GitHub.

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

Repository stars
674
Repository forks
189
7-day trend score
33

Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/automateyournetwork/netclaw/comfyui-topology-viz/badge)](https://verified-skill.com/skills/automateyournetwork/netclaw/comfyui-topology-viz)

Something wrong with this skill or its page? Tell us.