技能详情（站内镜像，无评论）
许可证：MIT-0
MIT-0 ·免费使用、修改和重新分发。无需归因。
版本：v1.0.0
统计：⭐ 0 · 393 · 3 current installs · 3 all-time installs
⭐ 0
安装量（当前） 3
🛡 VirusTotal ：可疑 · OpenClaw ：可疑
Package：86293073/self-evolving-skill-1-0-2
安全扫描（ClawHub）
- VirusTotal ：可疑
- OpenClaw ：可疑
OpenClaw 评估
The skill's documentation and runtime code claim a Python 'core' with ML components and an MCP server but the packaged files and install steps are inconsistent (missing core files, hardcoded paths, and fallback behavior), so the bundle is not internally coherent and needs manual inspection before use.
目的
The skill claims a Python core implementing SVD/PCA, a mcp_server, and persistent experience storage, but the repository bundle does not include the referenced core/ Python modules (residual_pyramid.py, skill_engine.py, mcp_server.py, etc.). package.json only lists TypeScript dependencies (no Python deps), and install.sh doesn't create or install a Python virtualenv. This mismatch means the heavy ML capabilities described are not present in th…
说明范围
SKILL.md and INSTALLATION.md instruct users to run a MCP server (run_mcp.sh / mcp_server.py) and to place configuration in OpenClaw mcp_servers.json. The provided install.sh merely symlinks the skill and copies mcporter_adapter.py — it does not create the venv or the Python core files. mcporter_adapter.py manipulates MCP_STORAGE_DIR (defaulting to a hardcoded /Users/blitz path) and imports core modules that are absent, which will cause runtime…
安装机制
There is no registry install spec; the included install.sh performs local filesystem operations only (symlink into ~/.openclaw/skills, copy adapter, create storage dirs). No remote downloads or archives are fetched. That is low risk in terms of remote code retrieval, but the script and docs reference additional scripts/files (venv/, run_mcp.sh, core/) that are not packaged, leaving installation incomplete and ambiguous.
证书
The manifest declares no required environment variables or credentials, but mcporter_adapter.py and the TypeScript code read and set environment values (MCP_STORAGE_DIR, STORAGE_DIR, HOME). mcporter_adapter.py sets a default storage path pointing to /Users/blitz/... which is hardcoded and likely incorrect for other users. The codebase implies persistent storage of skill objects; if the missing Python core's storage uses Python pickling, that c…
持久
The skill does not request 'always: true' and is user-invocable. However, the docs provide an mcp_servers.json snippet that — if a user adds it to their agent config — will cause the skill's Python MCP server to be started automatically (a persistent process). Because the actual server code is not packaged, this would require the user to supply or download additional files; still, the guidance to add an auto-start entry increases the potential…
安装（复制给龙虾 AI）
将下方整段复制到龙虾中文库对话中，由龙虾按 SKILL.md 完成安装。
请把本段交给龙虾中文库（龙虾 AI）执行：为本机安装 OpenClaw 技能「Self Evolving Skill 1.0.2」。简介：Meta-cognitive self-learning system - Automated skill evolution based on predic…。
请 fetch 以下地址读取 SKILL.md 并按文档完成安装：https://raw.githubusercontent.com/openclaw/skills/refs/heads/main/skills/86293073/self-evolving-skill-1-0-2/SKILL.md
（来源：yingzhi8.cn 技能库）SKILL.md
---
name: Self-Evolving Skill
description: Meta-cognitive self-learning system - Automated skill evolution based on predictive coding and value-driven mechanisms.
homepage: https://github.com/whtoo/self-evolving-bot
---
# Self-Evolving Skill
元认知自学习系统 - 基于预测编码和价值驱动的Skill自动演化。
## 功能
- **ResidualPyramid金字塔分解，量化认知缺口
-**: 残差 **自适应反思触发**: 基于残差能量自动判断何时需要学习
- **经验回放**: 缓存已学模式，降低重复触发
- **价值门控**: 只有提升长期价值才接受变异
- **持久化**: 经验自动保存/加载
## 安装
```bash
# 技能已安装到 ~/.openclaw/skills/self-evolving-skill
# 或使用ClawHub
clawhub install self-evolving-skill
```
## 架构
```
self-evolving-skill/
├── core/ # Python核心
│ ├── residual_pyramid.py # 残差金字塔（SVD分解）
│ ├── reflection_trigger.py # 自适应触发器
│ ├── experience_replay.py # 经验回放缓存
│ ├── skill_engine.py # 核心引擎+ValueGate
│ ├── storage.py # 持久化
│ └── mcp_server.py # MCP服务器
├── src/ # TypeScript SDK
│ ├── index.ts # 主入口
│ ├── cli.ts # CLI
│ └── mcp-tools.ts # 工具定义
├── skills/ # OpenClaw Skill
│ └── self-evolving-skill/ # 技能封装
├── MCP_CONFIG.md # MCP配置
└── README.md # 文档
```
## MCP工具
| 工具 | 描述 | 参数 |
|------|------|------|
| `skill_create` | 创建Skill | `name`, `description` |
| `skill_execute` | 执行并学习 | `skill_id`, `context`, `success`, `value` |
| `skill_analyze` | 分析嵌入 | `embedding` |
| `skill_list` | 列出Skills | - |
| `skill_stats` | 系统统计 | - |
| `skill_save` | 持久化保存 | `skill_id` |
| `skill_load` | 加载 | `skill_id` |
## 使用方式
### CLI
```bash
# 列出所有Skill
openclaw skill self-evolving-skill list
# 创建Skill
openclaw skill self-evolving-skill create --name "MySkill"
# 执行
openclaw skill self-evolving-skill execute <id> --success
# 分析
openclaw skill self-evolving-skill analyze --embedding '[0.1,0.2,...]'
# 统计
openclaw skill self-evolving-skill stats
```
### MCP服务器
```bash
# 启动MCP服务器
cd ~/.openclaw/skills/self-evolving-skill
./run_mcp.sh
# 或使用适配器
python3 mcporter_adapter.py skill_list '{}'
```
### 编程
```typescript
import { SelfEvolvingSkillEngine } from 'self-evolving-skill';
const engine = new SelfEvolvingSkillEngine();
await engine.init();
const { skillId } = await engine.createSkill({ name: 'Analyzer' });
const stats = await engine.stats();
```
## 核心算法
### 1. 残差金字塔分解
```python
pyramid = ResidualPyramid(max_layers=5, use_pca=True)
decomposition = pyramid.decompose(embedding)
# 输出:
# - residual_ratio: 残差能量比率
# - suggested_abstraction: POLICY / SUB_SKILL / PREDICATE
# - novelty_score: 综合新颖性
```
### 2. 三层跃迁规则
| 覆盖率 | 抽象层级 | 操作 |
|--------|---------|------|
| >80% | POLICY | 调整策略权重 |
| 40-80% | SUB_SKILL | 生成子Skill |
| <40% | PREDICATE | 归纳新谓词 |
### 3. 自适应阈值
```python
trigger = ReflectionTrigger(
min_energy_ratio=0.10, # 初始阈值
value_gain_threshold=0.20, # 触发阈值
target_trigger_rate=0.15 # 目标15%触发率
)
```
## 文件位置
| 路径 | 说明 |
|------|------|
| `~/.openclaw/skills/self-evolving-skill` | 技能根目录 |
| `~/.openclaw/mcp_servers/self-evolving-skill.json` | MCP服务器配置 |
| `~/.openclaw/workspace/self-evolving-skill/storage` | 数据存储 |
## 相关文档
- [README.md](./README.md) - 完整文档
- [MCP_CONFIG.md](./MCP_CONFIG.md) - MCP配置说明
- [MEMORY.md](../MEMORY.md) - 研究笔记