Skip to main content
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
clip skill Use CLIP for zero-shot image classification, image-text matching, and semantic image search without fine-tuning. CX Codex +1 1,100installs pytorch-fsdp2 skill Teaches a coding agent to wire PyTorch FSDP2 into a training loop: init, sharding, mixed precision and distributed checkpoints. CX Codex +1 1,100installs awq-quantization skill awq-quantization — a skill for 4-bit LLM weight quantization with AWQ: faster inference and lower GPU memory use. CX Codex +1 1,100installs axolotl skill Walks your agent through Axolotl training runs: YAML configs, LoRA and QLoRA, DPO, KTO, ORPO and GRPO methods, dataset formats. CX Codex +1 1,100installs huggingface-tokenizers skill Provides guides and code patterns for high-speed text tokenization using Hugging Face Tokenizers. Useful when building NLP pipelines and training custom vocabularies. CX Codex +1 1,100installs sentence-transformers skill sentence-transformers guides an agent through local text and image embeddings for RAG, semantic search, clustering, and similarity work. CX Codex +1 1,100installs hqq-quantization skill hqq-quantization is a skill for Claude Code and Codex that guides agents in quantizing LLM weights to 8/4/3/2/1 bits without calibration data, for serving with vLLM or Transformers. CX Codex +1 1,100installs verl-rl-training skill verl-rl-training provides step-by-step guidance for reinforcement learning post-training of LLMs using verl with FSDP, Megatron-LM, and Ray. CX Codex +1 1,100installs grpo-rl-training skill grpo-rl-training — a skill that walks your agent through GRPO fine-tuning with TRL: reward function templates and training configs. CX Codex +1 1,100installs quantizing-models-bitsandbytes skill quantizing-models-bitsandbytes quantizes LLMs to 8-bit or 4-bit via bitsandbytes to cut GPU memory use by 50% to 75%. Useful when fitting large models on limited VRAM or running QLoRA training. CX Codex +1 1,100installs optimizing-attention-flash skill optimizing-attention-flash — a skill that moves transformer attention onto Flash Attention for 2-4x speedup and 10-20x memory reduction. CX Codex +1 1,100installs fine-tuning-with-trl skill fine-tuning-with-trl — a skill for Claude Code, Cursor and other platforms adds LLM fine-tuning methods via TRL: SFT for instruction tuning, DPO for preference alignment, PPO and GRPO for reinforcement learning. CX Codex +1 1,100installs tensorboard skill This skill gives your agent working recipes for TensorBoard: log training metrics, inspect weight histograms, open the dashboard. Useful when comparing runs or hunting bottlenecks. CX Codex +1 1,100installs qdrant-vector-search skill This skill guides agents through Qdrant vector search for RAG and semantic search. Use it when your project needs metadata filters or scalable vector storage. CX Codex +1 1,100installs prompt-guard skill Integrates Meta's 86M parameter Prompt Guard model to detect jailbreaks and prompt injections in user prompts and RAG contexts before reaching the LLM. CX Codex +1 1,100installs unsloth skill Skill for Claude Code, Codex, Cursor and other platforms: helps set up fast fine-tuning via Unsloth library. Speeds up training 2–5×, cuts memory usage by 50–80%, supports LoRA and QLoRA. CX Codex +1 1,100installs segment-anything-model skill Covers Meta AI's Segment Anything Model for zero-shot image segmentation. Includes point, box and mask prompts, plus automatic mask generation for whole images. CX Codex +1 1,100installs gguf-quantization skill This skill guides llama.cpp GGUF quantization from 2 to 8 bits. It helps run LLMs locally on CPUs or Apple Silicon without dedicated GPU hardware. CX Codex +1 1,100installs fine-tuning-serving-openpi skill fine-tuning-serving-openpi is a Claude Code skill with workflows for fine-tuning and serving OpenPI models pi0, pi0-fast, and pi0.5. It guides setup, norm stats calculation, and policy inference server deployment. CX Codex +1 1,000installs fine-tuning-openvla-oft skill fine-tuning-openvla-oft guides fine-tuning and evaluation of OpenVLA-OFT robot action policies. Use it for LIBERO or ALOHA workflows with continuous action heads. CX Codex +1 1,000installs experiment-tracking-swanlab skill experiment-tracking-swanlab — a skill that shows your agent how to track ML runs in SwanLab: metrics, configs, media logging and a local dashboard. CX Codex +1 1,000installs evaluating-cosmos-policy skill evaluating-cosmos-policy — a skill for running NVIDIA Cosmos Policy evaluations on LIBERO and RoboCasa simulators and profiling inference latency. CX Codex +1 998installs evolving-ai-agents skill evolving-ai-agents — a skill for A-Evolve: it runs solve-observe-evolve cycles that rewrite an agent's prompts, skills and memory against benchmark scores. CX Codex +1 972installs ara-rigor-reviewer skill ara-rigor-reviewer reviews an ARA for epistemic rigor before publication or release. It examines evidence, claims, and research method after Level 1 validation. CX Codex +1 953installs ara-compiler skill ara-compiler turns papers, code, experiment logs, and notes into a structured research artifact with claims, evidence, an exploration graph, and code stubs. CX Codex +1 935installs ara-research-manager skill ara-research-manager — a skill that reviews the session after a task is done and writes decisions, experiments and dead ends into the ara/ directory. CX Codex +1 930installs

41–66 of 66

More in NeVibeCode

Project sponsors

Companies that keep this analytics open