Orchestra-Research/AI-Research-SKILLs

📦 🤖 Qoder agent support (npm) — #29

• ```bash
• npx @orchestra-research/ai-research-skills
• ```

📦 Notes

• No dependency changes; lockfile in sync, `npm audit` clean (0 vulnerabilities).
• Published via OIDC trusted publishing with Sigstore provenance.
• Thanks to @tianhao909 (Qoder) and @ADu2021 (coefficient tuning) for the contributions.
• Full Changelog: https://github.com/Orchestra-Research/AI-Research-SKILLs/compare/v1.7.1...v1.7.2

📦 🛡️ Drift guard

• New `scripts/check-inventory.sh` + `check-inventory.yml` CI job fails whenever documented counts diverge from the actual `SKILL.md` count on disk.

📦 📦 Marketplace sync hardening

• `sync-skills.yml` prunes build artifacts before zipping and fails loudly above 190 files instead of hitting the marketplace's 200-file rejection (#57).

🔒 🔒 Security

• > Note: v1.7.0 was tagged internally but never published — its publish failed on a lockfile desync introduced by the dep pins; v1.7.1 is the corrected, published release.
• 📦 npm: \`npm i -g @orchestra-research/ai-research-skills@1.7.1\` (published via OIDC trusted publishing with Sigstore provenance)

✨ What's New

• Autoresearch — the central orchestration skill that enables AI agents to autonomously conduct end-to-end research projects.

📦 Autoresearch Skill

• Two-loop architecture: inner loop (rapid experiments with measurable outcomes) + outer loop (synthesis, reflection, direction decisions)
• Routes to all 85 domain skills automatically — agents don't need to know which skill to use
• Continuous autonomous operation: mandatory `/loop 10m` (Claude Code) or cron job (OpenClaw)
• Research presentations: generates HTML/PDF progress reports with optimization trajectory plots
• Project memory: `findings.md` with "Lessons and Constraints" persists knowledge across sessions
• Structured workspace: research-state.yaml, findings.md, research-log.md, literature/, experiments/, src/, data/, to_human/
• Works with any compute: local GPU, cluster job submission, cloud, or CPU-only

📦 Demo Papers (produced by autoresearch)

• [Norm Heterogeneity → LoRA Brittleness](https://github.com/Orchestra-Research/AI-Research-SKILLs/tree/main/demos/autoresearch-norm-heterogeneity): Agent pivoted from null result on ETF overlaps to discovering norm CV as causal predictor (r=-0.99)
• [RL Algorithm Brain Scan](https://github.com/Orchestra-Research/AI-Research-SKILLs/tree/main/demos/autoresearch-rl-brain-scan): Agent found DPO is a rank-1 perturbation, orchestrating GRPO/TRL + SAELens/TransformerLens + paper writing

📦 Other Improvements

• WELCOME.md: cold-start bootstrap — one URL to go from zero to autonomous research
• Agent-friendly install path: `Read welcome.md and follow instructions` alongside `npx` command
• Windows support: symlink fallback to copy when Developer Mode unavailable
• Research Ideation (21-research-ideation) added to npm installer categories
• 86 total skills across 22 categories — complete research lifecycle coverage

📦 Install / Update

• ```bash
• npx @orchestra-research/ai-research-skills
• Read https://www.orchestra-research.com/ai-research-skills/welcome.md
• npx @orchestra-research/ai-research-skills install --all
• ```

✨ What's New

• Local project installation (`--local` flag): Install skills per-project instead of globally. Different projects can have different skill sets. Skills are copied into agent directories within the project, allowing them to be committed to version control for team sharing.
• Local agent path config: e.g., Cursor uses `.cursor/rules/` locally
• Lock file tracking: Track local installs via `.orchestra-skills.json`
• Interactive menu: "Install to project (local)" option added
• OIDC trusted publishing: Switched to pure OIDC auth for npm publish (no tokens needed), with `--provenance` for supply chain security

📦 Install / Update

• ```bash
• npx @orchestra-research/ai-research-skills
• ```

📦 Local Installation

• ```bash
• npx @orchestra-research/ai-research-skills install --local
• npx @orchestra-research/ai-research-skills list --local
• ```

✨ What's New

• OpenClaw support: Auto-detects and installs skills to `~/.openclaw/skills/`
• Shared `.agents/` directory: Installs to `~/.agents/skills/`, a common folder shared across many agents
• Now supports 8 agents total: Claude Code, OpenCode, OpenClaw, Cursor, Codex, Gemini CLI, Qwen Code, and shared `.agents/`

📦 Install / Update

• ```bash
• npx @orchestra-research/ai-research-skills
• ```

📦 OpenCode Support

• Added support for [OpenCode](https://opencode.ai) - the new AI coding assistant!
• ✅ Auto-detects OpenCode installations at `~/.config/opencode`
• ✅ Installs skills to OpenCode's skills directory
• ✅ Works alongside other agents (Claude Code, Cursor, etc.)

📋 Changes

• Added OpenCode to supported agents list
• Updated package to v1.1.0
• Cleaned up documentation to reflect actually supported agents
• Now supports 6 agents: Claude Code, OpenCode, Cursor, Codex, Gemini CLI, Qwen Code

📦 Installation

• ```bash
• npx @orchestra-research/ai-research-skills
• ```
• The installer will automatically detect OpenCode and install skills!
• ---
• Full Changelog: https://github.com/Orchestra-Research/AI-research-SKILLs/compare/v1.0.3...v1.1.0
• Closes #10

📦 🏢 Official Orchestra Research Organization

• Repository transferred to [Orchestra-Research](https://github.com/Orchestra-Research) organization
• All links updated to new org location
• GitHub automatically redirects old URLs

📦 🎬 Promo Video & Demo GIF

• Added animated demo GIF to README header
• Created Remotion-based promo video showcasing the installation flow
• Apple-style design with clean animations

📦 📦 npm Package v1.0.3

• Updated repository URL to Orchestra-Research org
• Package: `npx @orchestra-research/ai-research-skills`
• ---
• Full Changelog: https://github.com/Orchestra-Research/AI-research-SKILLs/compare/v0.14.0...v0.15.0

📦 📦 npm Package for All Coding Agents

• Install AI research skills to any coding agent with one command:
• ```bash
• npx @orchestra-research/ai-research-skills
• ```

📦 Supported Agents

• Claude Code
• Cursor
• Codex (OpenAI)
• Gemini CLI
• Qwen Code

✨ Features

• Interactive installer - Choose everything, quickstart, by category, or individual skills
• Auto-detection - Finds your installed coding agents automatically
• Update - Refresh only installed skills with latest versions
• Uninstall - Remove all or select specific skills
• ---

📦 npm Package

• `@orchestra-research/ai-research-skills` v1.0.2 on npm
• Skills stored in `~/.orchestra/skills/` with symlinks to each agent
• Works alongside Claude Code marketplace (alternative method)

📦 82 Total Skills (+1)

• TorchTitan added to Model Architecture - PyTorch-native distributed training for Llama 3.1
• Megatron-Core moved to Distributed Training category (now 6 skills)
• Post-Training confirmed at 8 skills: TRL, GRPO, OpenRLHF, SimPO, verl, slime, miles, torchforge

📝 Documentation

• README updated with npm as primary install method
• Claude Code marketplace kept as alternative
• All skill counts corrected
• ---

📦 📊 Stats

• | Metric | Value |
• |--------|-------|
• | Total Skills | 82 |
• | Categories | 20 |
• | Supported Agents | 5 |
• | npm Package | v1.0.2 |
• ---

📦 Quick Start

• ```bash
• npx @orchestra-research/ai-research-skills
• npx @orchestra-research/ai-research-skills list
• npx @orchestra-research/ai-research-skills update
• ```
• Full Changelog: https://github.com/zechenzhangAGI/AI-research-SKILLs/compare/v0.13.0...v0.14.0

📦 📝 ML Paper Writing Skill (20th Category!)

• Write publication-ready papers for top ML conferences by prompting your AI agent.
• Target NeurIPS, ICML, ICLR, ACL, AAAI, COLM
• Writing philosophy from Neel Nanda, Sebastian Farquhar, Gopen & Swan, Zachary Lipton, Ethan Perez
• Citation verification workflow - your agent will never hallucinate references
• LaTeX templates for 6 major conferences (ready to compile)
• 4 complete workflows: repo→paper, citations, templates, conference conversion

📦 🎪 Curated Demos Gallery

• See skills in action with real results.
• | Demo | What It Does |
• |------|--------------|
• | [NeMo Eval: GPQA Benchmark](https://github.com/zechenzhangAGI/Nemo-Eval-Skill-Demo) | Compare Llama 8B/70B/405B on graduate-level science |
• | [LoRA Without Regret](https://www.orchestra-research.com/perspectives/LLM-with-Orchestra) | Reproduce ML paper findings by just prompting an agent |
• ---

📦 📊 By the Numbers

• | Metric | Count |
• |--------|-------|
• | Total Skills | 77 |
• | Categories | 20 |
• | Lines of Documentation | ~120,000 |
• | LaTeX Templates | 6 conferences |

📦 🏆 Highlights

• Model Architecture: LitGPT, Mamba, NanoGPT, RWKV, Megatron-Core
• Fine-Tuning: Axolotl, LLaMA-Factory, Unsloth, PEFT
• Post-Training: TRL, GRPO, OpenRLHF, SimPO
• Inference: vLLM, TensorRT-LLM, llama.cpp, SGLang
• Interpretability: TransformerLens, SAELens, NNSight, pyvene
• And 50+ more...
• ---

📦 🚀 Quick Start

• ```bash
• /plugin marketplace add zechenzhangAGI/AI-research-SKILLs
• /plugin install fine-tuning@ai-research-skills
• /plugin install post-training@ai-research-skills
• /plugin install ml-paper-writing@ai-research-skills
• ```
• ---

📦 🔗 Links

• [📖 Full Documentation](https://github.com/zechenzhangAGI/AI-research-SKILLs)
• [🎪 Demos Gallery](https://github.com/zechenzhangAGI/AI-research-SKILLs/tree/main/demos)
• [🎥 Video: Reproduce Papers with AI Agents](https://www.youtube.com/watch?v=X0DoLYfXl5I)
• [📝 Blog: LLM Fine-Tuning with Orchestra](https://www.orchestra-research.com/perspectives/LLM-with-Orchestra)
• ---
• Star ⭐ the repo if you find it useful!

✨ 🆕 New Skills in This Release

• Lambda Labs - Reserved/on-demand GPU cloud with H100/A100, persistent filesystems, 1-Click Clusters
• Segment Anything (SAM) - Meta's zero-shot image segmentation with points/boxes/masks
• BLIP-2 - Vision-language pretraining with Q-Former, image captioning, VQA
• AudioCraft - Meta's MusicGen/AudioGen for text-to-music and text-to-sound

📦 📊 Library Stats

• 70 skills across 18 categories
• ~115,000 lines of documentation
• 58 gold-standard skills with comprehensive references
• Full AI research lifecycle coverage

📦 📦 Categories

• | Category | Count | Highlights |
• |----------|-------|------------|
• | Model Architecture | 5 | Megatron-Core, LitGPT, Mamba |
• | Fine-Tuning | 4 | Axolotl, PEFT, Unsloth |
• | Distributed Training | 5 | DeepSpeed, FSDP, Ray Train |
• | Infrastructure | 3 | Modal, SkyPilot, Lambda Labs |
• | Optimization | 6 | Flash Attention, GPTQ, AWQ |
• | Inference & Serving | 4 | vLLM, TensorRT-LLM, SGLang |
• + 4 more

📦 ⚠️ Pre-release Note

• This is a pre-release. Skills are functional but still undergoing battle-testing in production environments. Feedback welcome!
• Full Changelog: https://github.com/zechenzhangAGI/AI-research-SKILLs/commits/v0.10.0