AI-Research-SKILLs

general

AI-Research-SKILLs is a Claude Code skill that delivers a library of 98 specialized skills enabling AI agents to autonomously conduct the full AI research lifecycle — from ideation and literature survey through experiment execution to paper writing. The library spans 23 categories including model architecture, fine-tuning (Axolotl, Unsloth, PEFT, LLaMA-Factory), post-training (TRL, GRPO, verl), distributed training (DeepSpeed, Megatron-LM), safety & alignment (LlamaGuard, NeMo Guardrails), RAG, multimodal, mechanistic interpretability, and emerging techniques. Installation runs via `npx @orchestra-research/ai-research-skills`, auto-detecting agents like Claude Code, Cursor, and Gemini CLI and placing skills in `~/.orchestra/skills/`. The autoresearch orchestration skill ties everything together with a two-loop architecture that routes tasks to the appropriate domain skill. Designed for AI/ML researchers who want production-ready infrastructure knowledge — real code examples, troubleshooting guides, and battle-tested workflows — without spending time debugging tooling instead of testing hypotheses.