Skills for Real Engineers. Straight from my .agents directory.
39 skills
triton-sageattention
Agent skill from
Install Triton + SageAttention to accelerate ComfyUI (the sageattn attention_mode and inductor torch.compile used by WanVideoWrapper / many video graphs). Windows-first (triton-windows + woct0rdho prebuilt SageAttention wheels matched to torch/CUDA/python into the RIGHT python), plus Linux (official triton + build) and Mac (N/A → sdpa/MPS). Also covers the SAFE sdpa / no-compile fallback so an example that assumes sageattn + torch.compile still runs when these aren't installed (video-extend TRAP 5). Use when a loader crashes with "No module named 'sageattention'" or reports triton unavailable, when asked to speed up Wan/video workflows, or when deciding whether to install acceleration vs. fall back.
How to install triton-sageattention
install triton-sageattention
# install the comfyui-mcp pack with the skills CLI
npx skills add artokun/comfyui-mcp# or add only this skill to Claude Code: copy the folder
plugin/skills/triton-sageattention/
# into
~/.claude/skills/triton-sageattention/ (all projects)
.claude/skills/triton-sageattention/ (one project)
1# Triton + SageAttention (ComfyUI acceleration)23> See also [`comfyui-launch-flags`](../comfyui-launch-flags/SKILL.md) for the full4> attention / VRAM / cache flag matrix. Note the Z-Image exception: Z-Image is5> broken under `--use-sage-attention`, so launch it with6> `--use-pytorch-cross-attention` instead.78## Prefer kitchen INT8 attention when it is available910If `kitchen` action:"status" (or `panel_kitchen`) reports kitchen present and11`int8_attention_is_available` on this GPU, launch with **`--use-ck-attention`**12and **skip the sageattention wheel dance**. Kitchen INT8 attention is a ComfyUI13flag; it does not need a version-matched `sageattention` wheel. Restart14required, consent-gated like every restart.1516Only fall through to the Triton + SageAttention install below when kitchen INT817is **unknown** or **not available**. A failed kitchen probe is unknown, not a no.1819## Overview2021Two optional accelerators that many modern video graphs (especially kijai's22ComfyUI-WanVideoWrapper) reference by default:2324- SageAttention (`import sageattention`), a quantized attention kernel.25 Selected via a node's `attention_mode = sageattn` (WanVideoWrapper) or ComfyUI's26 `--use-sage-attention` startup flag. ~20 to 40% faster sampling on supported27 NVIDIA GPUs.28- Triton, the GPU kernel compiler that inductor `torch.compile` needs.29 WanVideoWrapper's `WanVideoTorchCompileSettings` (and any `torch.compile`/30 inductor node) compiles the model through Triton for another speedup.31…
Excerpt shown under the repository's MIT license.
More skills in comfyui-mcp
Keep exploring
Similar skill repositories
An agentic skills framework & software development methodology that works.
15 skills
Public repository for Agent Skills
20 skills
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
CCCU6 skills
Persistent Context Across Sessions for Every Agent – Captures everything your agent does during sessions, compresses it with AI, and injects relevant context back into future sessions. Works with Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot, OpenCode + More
CCCXGHGM+135 skills
Production-grade engineering skills for AI coding agents.
CCCXCU25 skills