NeVibeCode · repository · 66 items
ai-research-skills: a set by orchestra-research
The orchestra-research/ai-research-skills repository on GitHub: 66 cards with a description and an install command. They work in Claude Code, Codex, Cursor — 8 platforms in total.
About the repository
- GitHub
- orchestra-research/ai-research-skills
- Stars
- ★ 14,726
- Installs across cards
- 74,188
- License
- MIT
- Source updated
- 2026-10-11
- In the catalog since
- 2026-10-25
ml-paper-writing skill ml-paper-writing guides ML and AI paper drafting from a research repository, with literature checks and preparation for conference submissions.
CX Codex +1
2,600installs
creative-thinking-for-research skill creative-thinking-for-research — a skill that runs your research problem through eight cognitive-science frameworks for creativity, aimed at finding non-obvious directions in CS and AI.
CX Codex +1
1,400installs
academic-plotting skill academic-plotting turns a paper section or a results table into a figure: Gemini for architecture diagrams, matplotlib and seaborn for data charts.
CX Codex +1
1,400installs
brainstorming-research-ideas skill brainstorming-research-ideas — a skill for Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, and OpenCode that walks researchers through ten structured ideation frameworks.
CX Codex +1
1,400installs
autoresearch skill autoresearch manages autonomous research projects using a two-loop cycle for rapid experiment execution and periodic synthesis.
CX Codex +1
1,200installs
stable-diffusion-image-generation skill Walks your agent through Stable Diffusion image generation with HuggingFace Diffusers: text-to-image, img2img, inpainting, ControlNet.
CX Codex +1
1,200installs
systems-paper-writing skill systems-paper-writing is a Claude Code skill that outlines 10–12-page systems papers. It organizes page budgets and paragraph-level flows for submissions to OSDI, SOSP, ASPLOS, NSDI, and EuroSys.
CX Codex +1
1,100installs
ml-training-recipes skill ml-training-recipes is a Claude Code skill for setting up PyTorch neural-network training, with recipes for training loops, Muon and AdamW, mixed precision, and scaling laws.
CX Codex +1
1,100installs
presenting-conference-talks skill presenting-conference-talks generates Beamer LaTeX and PPTX presentation slides from a compiled research paper with speaker notes. Useful when preparing oral, spotlight, or invited conference talks.
CX Codex +1
1,100installs
lambda-labs-gpu-cloud skill The lambda-labs-gpu-cloud skill helps run Lambda Labs GPU instances for ML training and inference. Use it for SSH access, persistent storage, or multi-node clusters.
CX Codex +1
1,100installs
llava skill Wires LLaVA into your agent: ask questions about an image and keep the conversation going. Useful for image chat and document scans.
CX Codex +1
1,100installs
phoenix-observability skill phoenix-observability is a Claude Code skill for LLM tracing, evaluation, and monitoring. It activates while debugging LLM apps, testing datasets, or watching production systems.
CX Codex +1
1,100installs
pinecone skill A skill for working with Pinecone: create indexes, upsert vectors, filter by metadata, use namespaces and hybrid search. Useful for RAG and semantic search in production.
CX Codex +1
1,100installs
constitutional-ai skill constitutional-ai — a skill for training models to behave harmlessly through self-critique and AI feedback, with no human labels needed.
CX Codex +1
1,100installs
nemo-evaluator-sdk skill nemo-evaluator-sdk — a skill for Claude Code and five other agent environments. Runs language models through 100+ benchmarks from 18+ harnesses with containerized, multi-backend execution.
CX Codex +1
1,100installs
skypilot-multi-cloud-orchestration skill skypilot-multi-cloud-orchestration orchestrates ML training and batch workloads across 20+ clouds while cutting GPU expenses through spot auto-recovery and region selection.
CX Codex +1
1,100installs
rwkv-architecture skill Explains the RWKV architecture and shows how to run the model in Claude and RNN modes, train it, and handle long context.
CX Codex +1
1,100installs
blip-2-vision-language skill blip-2-vision-language — a Claude Code skill that teaches an agent to run Salesforce BLIP-2: image captioning, visual question answering, and multimodal chat. It bridges frozen image encoders to LLMs through a Q-Former.
CX Codex +1
1,100installs
autogpt-agents skill Guides AI coding assistants on building and running continuous AutoGPT agents with visual workflows. Useful for multi-step automations triggered by schedules or webhooks.
CX Codex +1
1,100installs
nemo-guardrails skill nemo-guardrails is a Claude Code skill that adds runtime checks for jailbreaks, PII, hallucinations, and toxicity before and after generation.
CX Codex +1
1,100installs
nnsight-remote-interpretability skill nnsight-remote-interpretability guides neural network interpretability and intervention using nnsight locally or remotely via NDIF. Useful for experimenting with 70B+ models without local GPU clusters.
CX Codex +1
1,100installs
llamaguard skill Wires Meta's 7-8B LlamaGuard model into your agent to filter user prompts and model responses across 6 safety categories.
CX Codex +1
1,100installs
torchforge-rl-training skill torchforge-rl-training is a skill for Claude Code that adds guidance for Meta’s torchforge reinforcement learning library, including GRPO training, reward functions and distributed runs.
CX Codex +1
1,100installs
miles-rl-training skill miles-rl-training — a skill that walks your agent through RL training of large MoE models on miles: FP8, INT4, speculative rollout. Useful when you're setting up training on H100/H200.
CX Codex +1
1,100installs
implementing-llms-litgpt skill implementing-llms-litgpt — a skill for Claude Code, Codex, Cursor and other agents that brings LitGPT into context: 20+ single-file model architectures (Llama, Gemma, Phi, Qwen, Mistral) with no abstraction layers.
CX Codex +1
1,100installs
mamba-architecture skill mamba-architecture — a skill for Claude Code, Codex, Cursor, GitHub Copilot, and Gemini CLI that gives your agent working knowledge of Mamba state-space models with O(n) complexity.
CX Codex +1
1,100installs
sparse-autoencoder-training skill sparse-autoencoder-training provides patterns to train and analyze Sparse Autoencoders using SAELens. Useful for finding interpretable features in transformer activations.
CX Codex +1
1,100installs
simpo-training skill simpo-training — a skill for Claude Code, Codex, Cursor, Copilot and Gemini CLI that trains language models via SimPO, a reference-free preference alignment method.
CX Codex +1
1,100installs
gptq skill gptq — a skill for post-training 4-bit quantization of LLMs: 4× memory reduction with under 2% perplexity loss.
CX Codex +1
1,100installs
slime-rl-training skill slime-rl-training guides LLM post-training with RL using the slime framework, connecting Megatron-LM and SGLang. Useful for GRPO workflows and tuning GLM, Qwen3, or DeepSeek models.
CX Codex +1
1,100installs
outlines skill outlines — a skill for generating valid JSON, XML and code from Pydantic schemas and grammars. Useful when a model has to return a fixed structure.
CX Codex +1
1,100installs
pyvene-interventions skill The pyvene-interventions skill guides causal intervention experiments on PyTorch models using pyvene. It covers causal tracing and activation patching on neural architectures.
CX Codex +1
1,100installs
sentencepiece skill A skill for Claude Code, Codex, Cursor, and other agents that trains and runs the SentencePiece tokenizer. It kicks in when a model needs a reproducible BPE or Unigram vocabulary built from raw text.
CX Codex +1
1,100installs
evaluating-code-models skill evaluating-code-models runs code-model benchmarks and reports pass@k metrics. Use it to compare models, test language coverage, or measure generated code quality.
CX Codex +1
1,100installs
openrlhf-training skill openrlhf-training — a skill that hands your agent ready commands for PPO, GRPO, RLOO and DPO runs on models from 7B to 70B+. Useful when training spans several GPUs or nodes.
CX Codex +1
1,100installs
distributed-llm-pretraining-torchtitan skill distributed-llm-pretraining-torchtitan configures distributed LLM pretraining using torchtitan with 4D parallelism (FSDP2, TP, PP, CP). Useful for scaling Llama 3.1 or DeepSeek V3 across 8 to 512+ GPUs.
CX Codex +1
1,100installs
transformer-lens-interpretability skill The transformer-lens-interpretability skill guides TransformerLens research on transformer internals, from activation caching and attention analysis to activation patching experiments.
CX Codex +1
1,100installs
modal-serverless-gpu skill modal-serverless-gpu — a skill for running ML workloads on Modal's serverless GPUs: training, inference, batch jobs. Useful when you'd rather not manage a cluster yourself.
CX Codex +1
1,100installs
sglang skill Equips your agent with patterns for serving LLMs via SGLang, configuring RadixAttention prefix caching, and decoding constrained JSON.
CX Codex +1
1,100installs
nanogpt skill nanogpt — a skill for Claude Code, Codex, Cursor and other platforms — educational GPT in 300 lines. Reproduces GPT-2 (124M) on OpenWebText, trains on Shakespeare in five minutes on CPU.
CX Codex +1
1,100installs
1–40 of 66