Self Evolving Agent
OpenClaw skill from miles990
Choose how to use this skill
You do not need every option. Choose the path your AI client supports. The stable page stays the same; versioned files are immutable.
1. Native installer
This listing has no registered native installer command. Use the complete package or source fallback below, depending on what your client supports.
Do not guess an installer command or replace an existing version without reviewing the diff.
2. Complete package recommended
Download the ZIP when available. It includes SKILL.md plus the references, security notes and version metadata.
No complete ProSkills package is published for this listing yet.3. Prompt-only
Copy the prompt above when the agent can read the stable page or when you want to adopt the workflow without installing a skill.
Need only the instruction file?
Download SKILL.md only if your client requires a single file. The complete ZIP is safer for a full installation because it preserves the references and release context.
No path installs or executes anything by itself. Your agent still needs access to the project files. Before updating, compare the installed version and review the diff.
// RATINGS
Not yet listed on ClawHub or SkillsMP
// README
Self-Evolving Agent
A Claude Code skill that enables autonomous goal achievement through iterative learning and self-improvement.
Quick Install | Documentation | Usage Manual | Examples | Troubleshooting
使用方式
/evolve [目標描述]
適用場景：
| 場景 | 說明 | 範例 |
|---|---|---|
| 🎯 複雜目標 | 多步驟、跨領域任務 | /evolve 建立完整的用戶認證系統 |
| 🔄 迭代開發 | 反覆調整直到達標 | /evolve 優化 API 回應時間降低 50% |
| 🧭 探索研究 | 方向不明確時 | /evolve 研究 codebase 架構 --explore |
| 💡 項目發想 | 從零開始構思 | /evolve 幫我發想 side project 功能 |
| 📝 制定規格 | 釐清需求 | /evolve 規劃訂單系統規格 |
| 📚 累積經驗 | 記錄解決方案 | /evolve 修復 bug 並記錄供未來參考 |
參數：
| 參數 | 說明 |
|---|---|
--explore | 探索模式 — 允許自主選擇方向 |
--emergence | 涌現模式 — 啟用跨領域連結 |
--autonomous | 自主模式 — 完全自主創新 |
--max-iterations N | 最大迭代次數（預設 10） |
--from-spec NAME | 從 spec-workflow 執行 |
運作原理
flowchart LR
subgraph 初始化
NS[🌟 北極星錨定]
PSB[規劃 → 設置 → 建構]
GOAL[目標分析<br/>深度訪談]
end
subgraph 執行循環
PDCA[計畫 → 執行 → 檢核 → 行動]
DIR[方向校正]
MEM[Git 記憶儲存]
end
NS --> PSB --> GOAL --> PDCA
PDCA --> DIR --> MEM
MEM -->|未完成| PDCA
MEM -->|完成| DONE[✅ 目標達成]
- 北極星錨定：任務開始前建立願景和完成標準，避免後續迷失方向
- 環境準備：檢查專案環境、安裝依賴、確認可以開始
- 目標分析：透過深度訪談釐清需求，分解成可執行的子任務
- PDCA 循環：反覆執行「計畫 → 執行 → 檢核 → 行動」直到完成
- 方向校正：每個里程碑後對照北極星，確保沒有偏離目標
- 記憶儲存：將學到的經驗存入 Git，供未來任務參考
💡 採用原子化架構，按需載入模組，Token 降低 92%。9 個強制檢查點確保行為穩定。
Quick Install
Plugin 版本（推薦） — 含 Hooks 自動提醒
# Step 1: 添加 marketplace
/plugin marketplace add miles990/self-evolving-agent
# Step 2: 安裝 plugin
/plugin install evolve@self-evolving-agent
更新到最新版本
/plugin marketplace update
/plugin update evolve@self-evolving-agent
One-line install（傳統 Skill 版本）
curl -fsSL https://raw.githubusercontent.com/miles990/self-evolving-agent/main/install.sh | bash
# Full install with hooks and memory
curl -fsSL https://raw.githubusercontent.com/miles990/self-evolving-agent/main/install.sh | bash -s -- --with-hooks --with-memory
Manual install
git clone https://github.com/miles990/self-evolving-agent.git
cp -r self-evolving-agent/skills /path/to/your/project/.claude/skills/evolve
核心哲學
給定目標，自主迭代直到完成。從失敗中學習，記憶經驗。北極星錨定，避免迷失方向。
- 目標導向：不是執行指令，而是達成目標
- 迭代改進：失敗是學習機會，每次嘗試都比上次更好
- 經驗累積：解決方案存入記憶，相同問題不再重蹈覆轍
模組文檔
| Module | Description |
|---|---|
| 00-getting-started | 入門與環境設定 |
| 01-core | 核心流程（PSB + PDCA） |
| 02-checkpoints | 強制檢查點（護欄） |
| 03-memory | 記憶系統操作 |
| 04-emergence | 涌現機制 |
| 05-integration | 外部工具整合 |
| 06-scaling | 大規模專案優化 |
| 99-evolution | 自我進化機制 |
相關資源
Related Projects
| Project | Description |
|---|---|
| claude-starter-kit | CLI 工具，快速初始化專案（內建此 skill） |
| claude-domain-skills | 16 non-technical domain skills |
| claude-software-skills | Software development skills |
References
- Reflexion Paper
- OpenAI Self-Evolving Agents Cookbook
- Andrew Ng - Agentic Design Patterns
- makepad-skills - Atomic architecture inspiration
📝 本專案也是經由 Self-Evolving Agent 迭代開發而成
License
MIT
// HOW IT'S BUILT
KEY FILES