@orchestra-research/ai-research-skills

Interactive UX Flow - One Command, Guided Experience

Step 1: Run Single Command

$ npx @orchestra-research/ai-research-skills

That's it. One command launches the full interactive experience.

Step 2: Orchestra Welcome Screen

  
   █████████  ██████████   █████████  █████   █████ ██████████  █████████  ███████████ ██████████    █████████
  ███░░░░░███░░███░░░░███ ███░░░░░███░░███   ░░███ ░░███░░░░░█ ███░░░░░███░█░░░███░░░█░░███░░░░███  ███░░░░░███
 ███     ░░░  ░███   ░░███░███    ░░░  ░███    ░███  ░███  █ ░ ░███    ░░░ ░   ░███  ░  ░███   ░░███░███    ░███
░███          ░███    ░███░░█████████  ░███████████  ░██████   ░░█████████     ░███     ░███    ░███░███████████
░███          ░███    ░███ ░░░░░░░░███ ░███░░░░░███  ░███░░█    ░░░░░░░░███    ░███     ░███    ░███░███░░░░░███
░░███     ███ ░███    ███  ███    ░███ ░███    ░███  ░███ ░   █ ███    ░███    ░███     ░███    ███ ░███    ░███
 ░░█████████  ██████████  ░░█████████  █████   █████ ██████████░░█████████     █████    ██████████  █████   █████
  ░░░░░░░░░  ░░░░░░░░░░    ░░░░░░░░░  ░░░░░   ░░░░░ ░░░░░░░░░░  ░░░░░░░░░     ░░░░░    ░░░░░░░░░░  ░░░░░   ░░░░░
  
                                    AI Research Skills

  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

                         Expert knowledge for AI research engineering
                            From model architecture to paper writing

                       82 skills  ·  20 categories  ·  7 agents

  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

                               Detecting your coding agents...
    
↓ Auto-detects after 1 second ↓

Step 3: Agent Detection Complete

  
   █████████  ██████████   █████████  █████   █████ ██████████  █████████  ███████████ ██████████    █████████
  ███░░░░░███░░███░░░░███ ███░░░░░███░░███   ░░███ ░░███░░░░░█ ███░░░░░███░█░░░███░░░█░░███░░░░███  ███░░░░░███
 ███     ░░░  ░███   ░░███░███    ░░░  ░███    ░███  ░███  █ ░ ░███    ░░░ ░   ░███  ░  ░███   ░░███░███    ░███
░███          ░███    ░███░░█████████  ░███████████  ░██████   ░░█████████     ░███     ░███    ░███░███████████
░███          ░███    ░███ ░░░░░░░░███ ░███░░░░░███  ░███░░█    ░░░░░░░░███    ░███     ░███    ░███░███░░░░░███
░░███     ███ ░███    ███  ███    ░███ ░███    ░███  ░███ ░   █ ███    ░███    ░███     ░███    ███ ░███    ░███
 ░░█████████  ██████████  ░░█████████  █████   █████ ██████████░░█████████     █████    ██████████  █████   █████
  ░░░░░░░░░  ░░░░░░░░░░    ░░░░░░░░░  ░░░░░   ░░░░░ ░░░░░░░░░░  ░░░░░░░░░     ░░░░░    ░░░░░░░░░░  ░░░░░   ░░░░░
  
                                    AI Research Skills

  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

  ✓ Found 3 coding agents:

       Claude Code      ~/.claude
       Cursor           ~/.cursor
       Gemini CLI       ~/.gemini

  Skills will be installed to all detected agents.

  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

                              Press Enter to continue...
    
↓ Auto-detects after 1 second ↓
  ╔═══════════════════════════════════════════════════════════════╗
                                                                 
     AI Research Skills                                          
     by Orchestra Research                                       
                                                                 
  ╚═══════════════════════════════════════════════════════════════╝

  ✓ Found 3 coding agents:

     Claude Code      ~/.claude
     Cursor           ~/.cursor
     Gemini CLI       ~/.gemini

  Skills will be installed to all 3 agents automatically.

  Press Enter to continue...
    

Step 4: Choose What to Install

  ? What would you like to install? (Use arrow keys)

     Everything                    All 82 skills (recommended for full setup)
      Select categories             Choose specific skill categories
      Select individual skills      Pick exactly which skills you need
      Quick start bundle            Popular skills for getting started (15 skills)
    

Step 5: Category Selection (if chosen)

  ? Select categories to install: (Space to select, Enter to confirm)

   Post-Training              8 skills   RLHF, GRPO, DPO, verl, slime...
   Inference Serving          4 skills   vLLM, TensorRT-LLM, SGLang...
   Model Architecture          6 skills   LitGPT, Mamba, TorchTitan...
   Fine-Tuning                 5 skills   Axolotl, Unsloth, PEFT...
   Distributed Training        6 skills   DeepSpeed, FSDP, Megatron...
   Optimization                6 skills   Flash Attention, GPTQ, AWQ...
   Mechanistic Interpretability4 skills   TransformerLens, SAELens...
   Data Processing             2 skills   NeMo Curator, Ray Data
   Safety & Alignment          3 skills   Constitutional AI, LlamaGuard...
   Infrastructure              3 skills   Modal, SkyPilot, Lambda Labs
   Evaluation                  3 skills   lm-eval-harness, BigCode...
   MLOps                       3 skills   W&B, MLflow, TensorBoard
   Agents                      4 skills   LangChain, LlamaIndex, CrewAI...
   RAG                         5 skills   Chroma, FAISS, Pinecone...
   Prompt Engineering          4 skills   DSPy, Instructor, Outlines...
   Observability               2 skills   LangSmith, Phoenix
   Multimodal                  7 skills   CLIP, Whisper, LLaVA...
   Emerging Techniques         6 skills   MoE, Model Merging, Pruning...
   Tokenization                2 skills   HuggingFace, SentencePiece
   ML Paper Writing            1 skill    NeurIPS/ICML paper writing

  ──────────────────────────────────────────────────────────────
  2 categories selected (12 skills)

  [Enter] Confirm   [Space] Toggle   [a] Select all   [n] Select none
    

Step 6: Confirm Installation

  ╔═══════════════════════════════════════════════════════════════╗
    Ready to Install                                             
  ╚═══════════════════════════════════════════════════════════════╝

  Skills to install:
    Post-Training (8)        verl, slime, miles, torchforge, grpo...
    Inference Serving (4)    vllm, tensorrt-llm, sglang, llama-cpp

  Target agents:
     Claude Code
     Cursor
     Gemini CLI

  Storage:
    Skills saved to:     ~/.agents/skills/
    Symlinks created in: ~/.claude/skills/, ~/.cursor/skills/, ~/.gemini/skills/

  ──────────────────────────────────────────────────────────────

  ? Proceed with installation?

     Yes, install 12 skills
      No, go back
    

Step 7: Installation (Animated Progress)

  Installing AI Research Skills...

  Downloading from GitHub...

     verl-rl-training          ━━━━━━━━━━ 12.1 KB
     slime-rl-training         ━━━━━━━━━━ 18.4 KB
     miles-rl-training         ━━━━━━━━━━  9.8 KB
     torchforge-rl-training    ━━━━━━━━━━ 11.2 KB
     grpo-rl-training          ━━━━━━━━━━ 15.3 KB
     trl-fine-tuning           ━━━━━━━━━━  8.7 KB
     openrlhf                   ━━━━━━━━━━  6.2 KB
     simpo                      ━━━━━━━━━━  4.1 KB
     vllm                       ━━━━━━━━━━ 14.5 KB
     tensorrt-llm               ━━━━━━━━━━ 11.8 KB
     sglang                     ━━━━━━━━━━  9.3 KB
     llama-cpp                  ━━━━━━━━━━  7.6 KB

  Creating symlinks...

     Claude Code     ~/.claude/skills/     12 skills
     Cursor          ~/.cursor/skills/     12 skills
     Gemini CLI      ~/.gemini/skills/     12 skills
    

Step 8: Success!

  
   █████████  █████  █████  █████████    █████████  ██████████  █████████   █████████
  ███░░░░░███░░███  ░░███  ███░░░░░███  ███░░░░░███░░███░░░░░█ ███░░░░░███ ███░░░░░███
 ░███    ░░░  ░███   ░███ ░███    ░░░  ░███    ░░░  ░███  █ ░ ░███    ░░░ ░███    ░░░
 ░░█████████  ░███   ░███ ░███         ░███         ░██████   ░░█████████ ░░█████████
  ░░░░░░░░███ ░███   ░███ ░███         ░███         ░███░░█    ░░░░░░░░███ ░░░░░░░░███
  ███    ░███ ░███   ░███ ░░███     ███░░███     ███░███ ░   █ ███    ░███ ███    ░███
 ░░█████████  ░░████████   ░░█████████  ░░█████████ ██████████░░█████████ ░░█████████
  ░░░░░░░░░    ░░░░░░░░     ░░░░░░░░░    ░░░░░░░░░ ░░░░░░░░░░  ░░░░░░░░░   ░░░░░░░░░
  
                               ✓ Installation Complete!

  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

  Installed 12 skills to 3 agents

  Your skills are ready! They'll automatically activate when relevant.

  Examples of what you can now do:

       Ask Claude/Cursor about GRPO training
       Get help setting up vLLM inference
       Learn verl for large-scale RL

  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

  Manage your skills anytime:

      npx @orchestra-research/ai-research-skills          Interactive menu
      npx @orchestra-research/ai-research-skills list     View installed
      npx @orchestra-research/ai-research-skills update   Update skills

  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

                 ★ Star us: github.com/orchestra-research/ai-research-skills
    

Alternative: Quick Commands (for power users)

Install Everything
$ npx @orchestra-research/ai-research-skills install --all
Skip interactive, install all 82 skills
Install Category
$ npx @orchestra-research/ai-research-skills install post-training
Install all skills in a category
Install Single Skill
$ npx @orchestra-research/ai-research-skills install verl
Install just one skill
Specific Agent Only
$ npx @orchestra-research/ai-research-skills install verl --agent claude
Install to just one agent

What Gets Created

~/.agents/                              # Canonical storage (single source)
├── .lock.json                          # Tracks versions & installed skills
└── skills/
    ├── 06-post-training/
    │   ├── verl/
    │   │   ├── SKILL.md                # Main skill instructions
    │   │   └── references/             # Detailed docs
    │   ├── slime/
    │   ├── miles/
    │   └── torchforge/
    └── 12-inference-serving/
        ├── vllm/
        └── sglang/

~/.claude/skills/                        # Symlinks (not copies)
├── verl → ~/.agents/skills/.../verl
├── slime → ~/.agents/skills/.../slime
└── vllm → ~/.agents/skills/.../vllm

~/.cursor/skills/                        # Same symlinks
└── (same links)

~/.gemini/skills/                        # Same symlinks
└── (same links)

Benefits:
   One copy of each skill, shared by all agents
   Update once, all agents get latest version
   Minimal disk space usage
    

Supported Agents (7 Verified)

All use the same SKILL.md format
Auto-detects which agents you have
Installs to all detected agents
Or specify with --agent claude
Skills activate when relevant

@orchestra-research/ai-research-skills

82 skills | 20 categories | 7 agents

npx @orchestra-research/ai-research-skills