AI-Research-SKILLs — Instalar & Demo ao vivo

Por que usar

Principais recursos

87 skills across 22 categories: model architectures, tokenization, fine-tuning, post-training, distributed training, optimization, inference, safety
Autoresearch orchestrator skill manages full experiment lifecycle
References folder per skill with official docs, GitHub issues, tutorials
Covers modern post-training (TRL, GRPO, SimPO, verl, slime) and inference (vLLM, SGLang, TensorRT-LLM, llama.cpp)

Demo ao vivo

Como fica na prática

ai-research-skill.replay ▶ pronto

0/0

Instalação

Repositório do autor · Documentação oficial de Skills do Claude

Casos de uso

Usos do mundo real: AI-Research-SKILLs

Fine-tune a Llama model with Unsloth + LoRA without Googling

👤 ML engineers running fine-tunes on a single GPU box ⏱ ~20 min intermediate

Quando usar: You have a dataset and want Claude to produce a runnable training script with correct flags.

Pré-requisitos

Unsloth installed — pip install unsloth — the skill will remind you of the nightly build caveats

Fluxo

Describe the target run

Use the unsloth skill. Fine-tune Llama-3-8B on my dataset at ~/data/train.jsonl, QLoRA, 3 epochs, output to ./out.✓ Copiado

→ Claude writes a script with correct Unsloth imports and real CLI flags
Ask for the config explanation

Walk me through each hyperparameter and why.✓ Copiado

→ Reasoning grounded in Unsloth's docs, not generic ML clichés

Resultado: A training script that runs on the first try.

Armadilhas

Claude mixes HF Trainer and Unsloth idioms — Insist on 'use the unsloth skill only' — don't pull from Transformers skill

Combine com: filesystem

Set up a vLLM server with correct tensor-parallelism for your GPUs

👤 Infra engineers serving LLMs in production ⏱ ~30 min advanced

Quando usar: You have 2-8 GPUs and want Claude to pick the right --tensor-parallel-size and --max-model-len.

Fluxo

State the hardware and model

Use the vllm skill. Serve Qwen2.5-72B on 4x H100s. Give me the exact launch command and sanity tests.✓ Copiado

→ Correct TP size and quantization recommendation
Ask for load-test script

Now give me a locust or vllm-benchmark script to verify throughput.✓ Copiado

→ Runnable benchmark using the right endpoint format

Resultado: A vLLM deployment with sanity checks and a benchmark baseline.

Armadilhas

Claude picks a TP size that doesn't divide the attention heads — The vLLM skill references list valid TP sizes per model family — have Claude cite that

Combine com: aws

Draft an ML paper with correct structure and citation conventions

👤 Researchers writing for NeurIPS/ICML/ICLR ⏱ ~60 min advanced

Quando usar: You have experiment results and need a paper outline that matches the venue's style.

Fluxo

Provide results and venue

Use the ml-paper-writing skill. Target ICLR. Here are my results [paste]. Draft the intro, method, and experiments sections.✓ Copiado

→ Structure follows venue conventions, ablation tables properly framed
Revise for reviewer concerns

What would Reviewer 2 push back on? Add preemptive responses.✓ Copiado

→ Concrete weaknesses, not generic 'add more experiments'

Resultado: A draft that passes initial sanity review and is worth polishing.

Armadilhas

Claude over-claims in the abstract — Explicitly tell it to mirror the restraint of top-tier accepted papers

Combine com: arxiv

Combinações

Combine com outros MCPs para 10× de alavancagem

ai-research-skill + filesystem

Store training configs, logs, and checkpoints while Claude drives experiments

Save the run config under experiments/<date>-<name>/ and tail the logs.✓ Copiado

ai-research-skill + arxiv

Pull related work while the ml-paper-writing skill drafts sections

Find 5 recent arXiv papers on GRPO and have the paper-writing skill weave them into the related work section.✓ Copiado

Ferramentas

O que este MCP expõe

Ferramenta	Entradas	Quando chamar
Autoresearch orchestrator	research goal	Complex multi-phase experiments
Fine-tuning (Axolotl, LLaMA-Factory, Unsloth, PEFT)	dataset + base model	SFT / LoRA / QLoRA runs
Post-training (TRL, GRPO, OpenRLHF, SimPO, verl, slime)	reward model or preference data	Alignment, preference optimization
Distributed training (DeepSpeed, FSDP, Megatron, Accelerate)	model + cluster topology	Multi-node or multi-GPU training
Inference (vLLM, SGLang, TensorRT-LLM, llama.cpp)	model + hardware	Serving a model efficiently
Optimization (Flash Attention, bitsandbytes, GPTQ, AWQ, GGUF)	model weights	Fitting a model on smaller hardware

Custo e limites

O que custa rodar

Cota de API: None for the skill itself
Tokens por chamada: Heavy — SKILL.md + references can load 5-10k tokens per sub-skill
Monetário: Free — skills are local files; you pay for compute you run
Dica: Scope prompts to one sub-skill at a time; the full 87-skill library is too big to load simultaneously.

Segurança

Permissões, segredos, alcance

Armazenamento de credenciais: No credentials — skill is prompts + references

Saída de dados: None from the skill itself

Some referenced frameworks have rapid API churn — confirm against the current version if a call fails.

Solução de problemas

Erros comuns e correções

Claude blends two frameworks (e.g. Accelerate + DeepSpeed) incorrectly

Name one skill in the prompt and tell Claude to ignore the other.

References feel outdated

The repo is TeX-heavy and may lag the bleeding edge. For brand-new frameworks, supplement with a fresh fetch of the official docs.

Skill not auto-invoking on a relevant prompt

Mention the framework by name — 87 skills overlap and auto-routing is fuzzy.

Alternativas

AI-Research-SKILLs vs. outros

Alternativa	Quando usar	Troca
scientific-agent-skill	You need biology/chemistry/clinical, not ML training infra	Different domain focus
huggingface MCP	You want live HF Hub operations rather than expert prompts	MCP gives you real API actions; this skill teaches the patterns

Mais

Recursos

📖 Leia o README oficial no GitHub

🐙 Ver issues abertas

🔍 Ver todos os 400+ servidores MCP e Skills