AI Research Skills: a Claude Code skill by Orchestra Research
For ML engineers who want working code patterns for fine-tuning, serving and evaluating models. Skip it if you only call hosted model APIs and never train or host models.
High-risk warning At least one scanner rated it high risk, so we keep it out of the top ten. Read its code before you install it.
- Score
- 68of 100
- Rank
- #191of 239
- Installs
- 50.1Kon skills.sh
- Stars
- 14.4Kon GitHub
- Security
- Warning 5 scanners
- Updated
- 2026
- Context
- ~82tokens always
Install
npx skills add https://github.com/Orchestra-Research/AI-Research-SKILLs
Run it in your terminal. View the source
Most people install it straight from GitHub, where it has 14.4K stars, so its skills.sh count understates how widely it is used.
Using the Claude app rather than Claude Code? You can add a skill there too. How to add it in the app
What does AI Research Skills do?
AI Research Skills is a set of 98 skills from Orchestra Research that act as reference guides for training, tuning, serving and evaluating AI models. Most members cover one library: axolotl, unsloth and trl for fine-tuning, deepspeed and pytorch-fsdp2 for distributed training, vllm, sglang and llama-cpp for inference, chroma, faiss and qdrant for retrieval, and agent frameworks like langchain and crewai. A few cover research itself, such as ml-paper-writing and brainstorming-research-ideas. The file we read for this entry is the PEFT member, which explains LoRA and QLoRA setup, rank and alpha choices, target modules per architecture, and how to load and merge adapters.
It comes as a set of skills that work together:
- 0-autoresearch-skill
- a-evolve
- academic-plotting (1.3K installs)
- accelerate
- audiocraft
- autogpt
- awq
- axolotl
- bigcode-evaluation-harness
- bitsandbytes
- blip-2
- brainstorming-research-ideas (1.3K installs)
When should you use AI Research Skills?
Use it when
- You need to fine-tune a 7B to 70B model on limited GPUs
- You are choosing an inference server such as vLLM
- You are setting up model evaluation or experiment tracking
Skip it when
- You only use hosted chat APIs
- You have no GPU access or training budget
What it needs and what it can touch
You need
- Python with the libraries each skill covers, such as peft and transformers
- GPU hardware for training work
What it can do
- Suggests pip installs and Python training code for Claude or you to run
From reading its instructions, not from running it.
Why it scores 68
Strongest on quality (100 of 100), weakest on adoption (52).
Is AI Research Skills safe to install?
At least one scanner rated it high risk, so we keep it out of the top ten. Read its code before you install it. Scanners often miss what other scanners catch, and a skill can change after a scan. Read its source before you install it.
| Scanner | Result | Date |
|---|---|---|
| Gen Agent Trust Hub | Passed | Sep 15, 2026 |
| Socket | Medium risk | Sep 15, 2026 |
| Snyk | Low risk | Sep 15, 2026 |
| Runlayer | High risk | Mar 6, 2026 |
| ZeroLeaks | Medium risk | Apr 16, 2026 |
Details
- Installs
- 50,073
- Repository stars
- 14,431
- Last repository change
- Tokens always loaded
- about 82
- Tokens when used
- about 2,898
- License
- MIT
- Data confidence
- High
Did it do the job for you?
AI Research Skills
Get AI Research Skills