Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add artokun/comfyui-mcp --skill train-character-loragit clone --depth 1 https://github.com/artokun/comfyui-mcpWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/artokun/comfyui-mcp/train-character-lora)<a href="https://agentmods.dev/skills/artokun/comfyui-mcp/train-character-lora"><img src="https://agentmods.dev/badge/skills/artokun/comfyui-mcp/train-character-lora/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/artokun/comfyui-mcp/train-character-lora"><img src="https://agentmods.dev/badge/skills/artokun/comfyui-mcp/train-character-lora.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00097 | $0.01406 |
| Opus 5 | $0.00048 | $0.00703 |
| Sonnet 5 | $0.00019 | $0.00281 |
| Haiku 4.5 | $0.00010 | $0.00141 |
Grade A, and why
train-character-lora scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 12d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 95 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Train a Character LoRA (local, Flux.1-dev)
Overview
The trainer runs ostris ai-toolkit's run.py inside a headless GPU Docker container,
driven through the three train_* MCP tools. You (the LLM) are the UI. Each takes
an action: train_prepare_dataset owns the datasets, train_start owns the jobs, and
train_doctor owns the trainer itself. You generate the dataset, launch the job, watch
progress, and the finished LoRA lands in ComfyUI models/loras/ and the LoRA catalog
without further steps.
- Base model: FLUX.1-dev (the best proven character consistency; needs ~24GB VRAM with quantization, RTX 4090 class).
- Phase-1 scope: character LoRAs only. Style/slider/edit and other bases come later.
The flow (tool sequence)
train_doctor {action:"doctor"}. Preflight once per session. Checks docker daemon,--gpus allGPU passthrough, trainer image, HF_TOKEN. Ifimage:false, runtrain_doctor {action:"build_image"}(one-time, several minutes, since it builds CUDA plus torch plus ai-toolkit). IfhfTokenSet:false, warn the user: the first run downloads FLUX.1-dev (gated HF repo) and needsHF_TOKENin the MCP server env.train_prepare_dataset {action:"prepare"}. Stage the images. See "Dataset" below.train_start {action:"start"}. Launch. Returns a job id at once; training runs detached.train_start {action:"status", id}. Poll progress (progress.step/totalSteps/loss, recentsamples,logtail). Poll on a slow cadence (every few minutes). A 2000-step run is roughly an hour on a 4090. Don't block on it.- Done.
status:"completed"means the.safetensorswas copied tomodels/loras/<name>.safetensorsand upserted into the LoRA catalog (resulthas the paths and catalog id). Verify by loading it in a Flux workflow (LoraLoaderModelOnly, strength 1.0) with the trigger word in the prompt.
Dataset guidance
Call train_prepare_dataset {action:"prepare"} with name, items: [{path, caption?}, ...]
and a defaultCaption.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 12d ago First seen · 95 lines · 97 tokens per session scan A 76e6643f9482
train-character-lora is a skill published in the GitHub repository artokun/comfyui-mcp (740 stars, last pushed 2d ago), licensed MIT. It adds 97 tokens to every session and 1,406 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
minimax-h3
Use when writing or debugging prompts for MiniMax H3 (Hailuo 3) video-with-audio generation, running the open weights locally in ComfyUI, choosing a quant or an acceleration LoRA for the VRAM you have, wiring reference-to-video with images, video or audio, or when a generated clip produces gibberish speech, drifts off…
seedance
Use when writing or debugging prompts for ByteDance Seedance video models (Seedance 2.5, 2.0, 2.0 Mini, 1.5 Pro, 1.0) on Dreamina, Jimeng AI, Doubao, BytePlus ModelArk or ComfyUI, when a generated video drifts off the reference face, grows unwanted subtitles or watermarks, duplicates a character, jumps at an extension…
Prompt craft for ComfyUI generation
Use when writing prompts for ComfyUI image or video models: choosing tags vs literary format, fixing composition and anatomy artifacts, picking samplers per LoRA, and writing short movement prompts for video. Neutral examples only.
emotion-to-camera-language
A guide for turning vague visual feelings—such as cinematic, atmospheric, elegant, or healing—into concrete image or video prompt details. It uses lighting direction, depth of field, camera position, and the subject’s state.
comfyui-lora-training
Prepare datasets and configure LoRA training for character consistency. Covers FLUX (AI-Toolkit, SimpleTuner, FluxGym) and SDXL (Kohyass) training with step-by-step guidance. Use when training custom character LoRAs.
comfyui-prompt-engineer
Craft model-specific prompts optimized for the target checkpoint and identity method. Handles FLUX, SDXL, SD1.5, and Wan video models with proper syntax, quality tags, and negative prompts. Use when generating or refining prompts for ComfyUI workflows.