AGI-Distiller
Health Warn
- License — License: NOASSERTION
- Description — Repository has a description
- Active repo — Last push 0 days ago
- Low visibility — Only 6 GitHub stars
Code Pass
- Code scan — Scanned 1 files during light audit, no dangerous patterns found
Permissions Pass
- Permissions — No dangerous permissions requested
No AI report is available for this listing yet.
Not just another skill collection. A living knowledge distillation system that self-evolves by reading technical content. 不只是 skill 集合,是一个能自我进化的知识蒸馏系统。
🧪 AGI Distiller
版权声明 Copyright:© 2026 All Rights Reserved. 未经作者书面许可,禁止复制、修改、分发、商用、用于各类学科竞赛。
Knowledge Distillation System for AI Coding Agents
Read. Distill. Evolve. Repeat.
🌟 Vision
AGI Distiller is not just another skill collection. It is a living knowledge distillation system that:
- Reads high-quality technical content (articles, blogs, papers, WeChat posts)
- Extracts pain points, solutions, and actionable rules
- Precipitates them into cross-platform agent skills, behavioral rules, and collaboration protocols
- Evolves itself — the more it reads, the more capable it becomes
The ultimate goal: An AI agent that continuously self-improves by reading, distilling, and applying knowledge — one article at a time.
🔬 What Makes This Different
| Dimension | Traditional Skill Repos | AGI Distiller |
|---|---|---|
| Source | Developer experience | High-quality technical content |
| Creation | Hand-written | Distilled from content |
| Evolution | Manual PRs | Self-evolving |
| Focus | General coding | Terminal AI agent efficiency |
| Language | English only | Bilingual (EN + CN) |
| Philosophy | "Here are the skills" | "Here is the distillery that produces skills" |
🏗️ System Architecture
┌─────────────────────────────┐
│ Knowledge Sources │
│ laodad.com · WeChat · Blogs │
│ arXiv · GitHub Issues · More │
└─────────────┬───────────────┘
│ Read
┌─────────────▼───────────────┐
│ Distillation Pipeline │
│ │
│ ① Extract Pain Points │
│ ② Extract Solutions │
│ ③ Derive Actionable Rules │
│ ④ Precipitate into Skills │
│ ⑤ Update Behavior Rules │
└─────────────┬───────────────┘
│ Output
┌─────────────────────────┼─────────────────────────┐
▼ ▼ ▼
┌───────────────┐ ┌───────────────┐ ┌───────────────┐
│ Agent Skills │ │ Rules & Specs │ │ Memory & │
│ (SKILL.md) │ │ (ATOMCODE.md) │ │ Knowledge │
│ Cross-platform│ │ Best practices│ │ (notes/) │
└───────────────┘ └───────────────┘ └───────────────┘
📦 What's Inside
Agent Spec Files (OpenAI Open Standard)
| File | Purpose | Size |
|---|---|---|
SOUL.md |
Agent personality (name, character, service targets) | ~905 chars |
AGENTS.md |
Permission matrix + red line rules (core security) | ~2016 chars |
USER.md |
Current user identity and role | ~359 chars |
TOOLS.md |
Tool usage guide | ~1013 chars |
IDENTITY.md |
Agent identity metadata | ~753 chars |
HEARTBEAT.md |
Session / health tracking | ~235 chars |
| Total | ~5281 chars |
Skills (Cross-Platform)
| Skill | Description | Status |
|---|---|---|
acceptance-checker |
7-item acceptance checklist for every task | ✅ Live |
debug-flow |
5-step debugging workflow (reproduce → locate → fix → test → regression) | ✅ Live |
deploy-checker |
8-item pre-release checklist (scope, diff, test, UI, rollback, notes, observe, honesty) | ✅ Live |
task-automator |
Automation task writer — write repeatable, verifiable automation workflows | ✅ Live |
task-briefer |
Structured task brief template (background, goal, scope, limits) | 🚧 Next |
code-review-p0 |
P0/P1/P2 graded code review | 🚧 Next |
cli-safety |
Agent-friendly CLI execution rules | 🚧 Next |
Knowledge Base
sources/— Index of all distilled content with extraction metadatarules/ATOMCODE.md— Comprehensive behavioral specification (12 sections)DISTILLER.md— Distillation pipeline specificationROADMAP.md— 4-phase development roadmap
🔧 Installation
Claude Code
/plugin marketplace add agi-distiller/agi-distiller
/plugin install agi-distiller@agi-distiller
Codex CLI
npx skills add agi-distiller/agi-distiller
Manual (Any Agent)
git clone https://github.com/TrueFurina/AGI-Distiller.git
cp -r AGI-Distiller/skills/* ~/.claude/skills/
Platform Compatibility
| Platform | Path | Status |
|---|---|---|
| Claude Code | ~/.claude/skills/ |
✅ |
| Codex CLI | ~/.codex/skills/ |
✅ |
| Cursor | .cursor/skills/ |
✅ |
| Gemini CLI | .gemini/skills/ |
✅ |
| GitHub Copilot | .github/skills/ |
✅ |
| OpenCode | ~/.config/opencode/skills/ |
✅ |
| Windsurf | .windsurf/skills/ |
✅ |
🧪 The Distillation Pipeline
Every skill in this repository is born from a structured process:
1. READ an article
2. EXTRACT:
- Pain point: What problem does this solve?
- Solution: How does it solve it?
- Actionable rule: What should the agent do differently?
- Trap: What should the agent avoid?
3. CLASSIFY: Does this fit an existing skill? Or need a new one?
4. PRECIPITATE:
- If new rule → update ATOMCODE.md
- If cross-session value → write to memory
- If high-frequency pattern → create/update skill
5. VERIFY: Is the skill usable? Does it solve the original pain point?
See DISTILLER.md for the complete pipeline specification.
🗺️ Roadmap
Phase 1: Foundation (Current)
- Core distillation pipeline design
- 4 production skills
- 12-section behavioral specification (ATOMCODE.md)
- 14 persistent memory entries
- 22 articles distilled from laodad.com
- GitHub repository live
Phase 2: Growth (Next 30 days)
- 20 skills total
- CI pipeline
- Marketplace registration
- Automated distillation pipeline
- Community contributions
Phase 3: Ecosystem (3 months)
- 50+ skills
- Multi-source distillation (laodad.com, WeChat, Medium, arXiv)
- Web catalog for browsing skills
- 10,000+ GitHub stars
🤝 Contributing
We welcome contributions of all kinds:
- 📝 New sources: Suggest a high-quality article to distill
- 🔧 New skills: Submit a skill based on the distillation template
- 🌐 Translations: Help translate skills to other languages
- 🐛 Bug fixes: Improve existing skills
See CONTRIBUTING.md for guidelines.
📚 Sources Distilled
| Source | Articles | Category | Status |
|---|---|---|---|
| laodad.com | 22 | AI programming efficiency | ✅ |
| 卡码大模型 (WeChat) | 1 | CLI & Agent | ✅ |
| More coming... | 🚧 |
📄 License
MIT — use these skills in your projects, teams, and tools. Full credit to the original authors of the articles that inspired each skill.
⭐ Star History
If this project resonates with you, give it a star ⭐ — it helps more people find this vision of self-evolving AI agents.
"The endgame is not CLI. The endgame is every piece of software growing a set of interfaces that an Agent can call, verify, audit, and be safely confined by. CLI is just the first one ready."
— 卡码大模型
License & Usage Notice
Source-Available · All Rights Reserved
This project is source-available and all rights are reserved by the author. The code is provided for viewing and evaluation purposes only — access does not grant any right to copy, modify, redistribute, use commercially, or create derivative works. Unauthorized reuse may carry legal risk. Contact the author for explicit written permission before any other use.
本仓库为「源码可查、权利保留」项目(source-available / all-rights-reserved)。代码仅供查看与评估,未授权任何复制、再分发、修改、商用或衍生创作。擅自借用代码存在法律风险;如有需要请先联系作者获取明确书面许可。
Reviews (0)
Sign in to leave a review.
Leave a reviewNo results found