- Home
- Skills
- Testing & Debugging
- ComfyUI AMD WSL2 Doctor
ComfyUI AMD WSL2 Doctor
Diagnoses and fixes broken ComfyUI installations running on AMD Radeon GPUs under WSL2 (ROCm, ZLUDA, DirectML).
$50
Works with the AI Agents you already use
ComfyUI AMD WSL2 Doctor
Example session with this skill installed
My ComfyUI was working yesterday, but now I get 'RuntimeError: Found no NVIDIA driver' even though I'm using an AMD 7900XTX on WSL2. Help.
- Read your context and instructions
- Compiled the comfyui amd wsl2
The WSL2 passthrough layer is likely broken or the ROCm kernel isn't visible to PyTorch. Run this to check visibility: python3 -c "import torch; print(torch.cuda.is_available())" If it returns False, paste the output of 'ls -l /dev/kfd' so I can check your hardware permissions.
Connects securely to your tools. The creator never sees your data.
About this skill
To start, simply import this skill package into your AI agent environment and instruct your agent:
"Load the skill defined in SKILL.md and help me diagnose my ComfyUI AMD setup."
The agent will actively guide you to isolate which layer is failing. When troubleshooting, it will prompt you:
- To paste the exact traceback or describe the symptom (e.g. black images, OOM crash, GPU not found).
- To run targeted diagnostic commands (one at a time) to identify the root cause.
- To confirm before executing any package reinstallation or modification.
---
About ComfyUI AMD WSL2 Doctor
ComfyUI AMD WSL2 Doctor is a production-grade diagnostic skill package designed to help developers, creators, and AI artists running ComfyUI on AMD Radeon GPUs inside WSL2 (Windows Subsystem for Linux).
AMD Radeon support under WSL2 is highly fragile because it sits at the intersection of Windows GPU drivers, WSL2 passthrough layers, ROCm/HIP runtimes, and PyTorch ROCm wheels. A single update to any of these components can silently break your entire image generation pipeline.
This skill provides your AI agent with a specialized, step-by-step diagnostic framework. It avoids broad "reinstall everything" suggestions and instead isolates the exact failure point across your system layers within minutes.
Key Highlights:
- Precise Layer Isolation: Determines whether the failure lies in the Windows WDDM driver, WSL2's DXG bridge, system ROCm libraries, PyTorch wheels, or custom node dependencies.
- Custom Node Protection: Provides strategies to pin PyTorch version using constraints, preventing ComfyUI Manager or custom node installers from overwriting your ROCm build with a CPU or CUDA version.
- Numerical Precision Fixing: Addresses AMD-specific NaN/black image bugs caused by bfloat16 to numpy conversions and recommends targeted FP32 workarounds.
- VRAM & Memory Tuning: Solves WSL2 RAM caps and PyTorch allocator limits that trigger Out-of-Memory crashes even on high-VRAM setups.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Testing & Debugging.
Requires shell access to WSL2. Version-specific commands (ROCm/PyTorch) may require verification against current release cycles. Not for NVIDIA/CUDA hardware.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 26 days ago
- Software
Python
Permissions
Requires file reading to access skill guides, and network/browser access to cross-reference AMD compatibility pages and PyTorch wheel releases for the user's specific GPU and driver version.