← Back to Skills Marketplace
pagoda111king

Autonomous Learning Cycle

by pagoda111king · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
133
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install autonomous-learning-cycle
Description
实现每17分钟自主执行任务、提取模式、评估自信、自动创建技能并生成每日/每周反思与新学习方向的闭环系统。
README (SKILL.md)

🔄 Autonomous Learning Cycle - 自主学习循环系统

版本: 1.0.0
作者: AIGC Autonomous Evolution System
描述: 完整的 17 分钟自主进化循环系统,实现从任务执行到自主进化的闭环 发布日期: 2026-03-30 安装量: 0 (v1.0.0 首发) 社区: 第一个完整的自主进化系统 对标: pskoett/self-improving-agent (6.2K installs)


🎯 系统能力

本 Skill 提供一个完整的17 分钟自主进化循环系统,包含:

6 大核心引擎

引擎 功能 文件
进化引擎 任务选择 + 自主执行 engines/evolution-engine.js
学习提取器 从任务提取可复用模式 engines/extractor.js
自信度评估 计算模式可靠性评分 engines/confidence.js
技能创建器 高自信模式→可复用技能 engines/skill-creator.js
反思引擎 每日/每周反思报告 engines/reflection.js
学习方向生成 自主发现新方向 + 生成任务 engines/learning-direction.js

4 大定时任务

任务 周期 说明
自主进化循环 */17 * * * * 每 17 分钟执行一轮学习循环
每日反思 0 23 * * * 每日 23:00 生成反思报告
每周反思 0 20 * * 0 每周日 20:00 生成周反思
学习方向生成 0 6 * * * 每日 06:00 生成新学习方向

完整工作流

定时触发 (17 分钟)
    ↓
自主选择任务 (按优先级 + 依赖)
    ↓
执行任务
    ↓
提取学习模式
    ↓
评估自信度
    ↓
高自信 → 自动创建技能
    ↓
每日反思 (23:00) → 识别改进点
    ↓
学习方向生成 (06:00) → 发现新技能 + 生成新任务
    ↓
等待下一轮 (17 分钟)

🚀 快速开始

安装

方法 1: ClawHub (推荐)

clawhub install autonomous-learning-cycle

方法 2: 手动安装

# 克隆或复制此 skill 到你的技能目录
cp -r skills/autonomous-learning-cycle ~/.jvs/.openclaw/skills/

方法 3: Git 安装

git clone https://github.com/YOUR_USERNAME/autonomous-learning-cycle.git ~/.jvs/.openclaw/skills/autonomous-learning-cycle

初始化

# 1. 创建必要目录
node skills/autonomous-learning-cycle/init.js

# 2. 设置定时任务
node skills/autonomous-learning-cycle/setup-cron.js

# 3. 启动系统
node skills/autonomous-learning-cycle/start.js

手动控制

# 执行一轮循环
node engines/evolution-engine.js run

# 查看系统状态
node engines/evolution-engine.js status

# 生成反思报告
node engines/reflection.js daily
node engines/reflection.js weekly

# 生成学习方向
node engines/learning-direction.js auto

📁 文件结构

autonomous-learning-cycle/
├── SKILL.md                    # 本文件
├── init.js                     # 初始化脚本
├── setup-cron.js               # Cron 设置脚本
├── start.js                    # 启动脚本
├── engines/                    # 核心引擎
│   ├── evolution-engine.js     # 进化引擎
│   ├── extractor.js            # 学习提取器
│   ├── confidence.js           # 自信度评估
│   ├── skill-creator.js        # 技能创建器
│   ├── reflection.js           # 反思引擎
│   └── learning-direction.js   # 学习方向生成
├── handlers/                   # Hook 处理器
│   ├── session-start.js        # Session 启动处理
│   └── file-generated.js       # 文件生成处理
├── configs/                    # 配置文件
│   ├── cron-jobs.json          # Cron 任务配置
│   └── confidence-config.json  # 自信度配置
└── docs/                       # 文档
    ├── INSTALL.md              # 安装指南
    ├── USAGE.md                # 使用指南
    └── ARCHITECTURE.md         # 架构文档

🎯 适用场景

适合使用本 Skill 的场景

  • ✅ 你想要一个自主学习和进化的 AI 系统
  • ✅ 你希望 AI 能够自主发现能力差距
  • ✅ 你想要每日/每周自动反思
  • ✅ 你希望从经验中提取可复用模式
  • ✅ 你想要自动创建技能
  • ✅ 你希望系统无需人工干预持续进化

不适合使用本 Skill 的场景

  • ❌ 你只需要简单的定时任务(用基础 cron 即可)
  • ❌ 你不需要学习提取和反思功能
  • ❌ 你希望完全手动控制每个任务

📊 系统指标

性能指标

指标 数值
单轮循环时间 17 分钟
单次任务执行 2-15 分钟
反思报告生成 \x3C5 秒
学习方向生成 \x3C30 秒
内存占用 \x3C50MB

进化指标

指标 目标 说明
每日循环数 84 轮 24 小时 × 3.5 轮/小时
每日完成任务 20+ 平均每轮 0.25 个任务
每日提取模式 20+ 每个任务提取 1 个模式
每周创建技能 5+ 高自信模式转化
自信度增长 +0.1/轮 重复使用提升可靠性

🔧 配置选项

自信度配置 (configs/confidence-config.json)

{
  "weights": {
    "baseScore": 0.5,
    "successRate": 0.3,
    "usageBonus": 0.15,
    "timeDecay": 0.05,
    "qualityBonus": 0.1
  },
  "thresholds": {
    "high": 0.7,
    "medium": 0.4,
    "low": 0.0
  },
  "decay": {
    "daysToHalf": 30,
    "minDecay": 0.5
  }
}

Cron 配置 (configs/cron-jobs.json)

{
  "jobs": [
    {
      "name": "自主进化循环",
      "schedule": "*/17 * * * *",
      "command": "node engines/evolution-engine.js run"
    },
    {
      "name": "每日反思",
      "schedule": "0 23 * * *",
      "command": "node engines/reflection.js daily"
    },
    {
      "name": "每周反思",
      "schedule": "0 20 * * 0",
      "command": "node engines/reflection.js weekly"
    },
    {
      "name": "学习方向生成",
      "schedule": "0 6 * * *",
      "command": "node engines/learning-direction.js auto"
    }
  ]
}

📈 使用案例

案例 1: 编程能力提升

场景: 你想要提升编程能力,成为大师级开发者

配置:

  • 任务队列:编程相关任务(实现功能、修复 bug、学习新技术)
  • 技能发现:编程相关技能(algorithm, debugging, testing)
  • 反思重点:代码质量、测试覆盖率、性能优化

预期效果:

  • 每日完成 5-10 个编程任务
  • 每周创建 2-3 个编程技能
  • 30 天后编程能力显著提升

案例 2: 学习新领域

场景: 你想要学习新领域(如 AI/ML、区块链、Web3)

配置:

  • 任务队列:学习任务(阅读文档、实践项目、写总结)
  • 技能发现:新领域相关技能
  • 反思重点:理解深度、实践能力、知识体系

预期效果:

  • 每日学习 3-5 个新概念
  • 每周创建 1-2 个领域技能
  • 30 天后掌握新领域基础

案例 3: 团队项目管理

场景: 你想要管理复杂项目,协调多任务并行

配置:

  • 任务队列:项目任务(分解、分配、追踪)
  • 技能发现:项目管理相关技能
  • 反思重点:进度、质量、风险

预期效果:

  • 自主追踪项目进度
  • 自动识别风险并预警
  • 生成项目报告

🎯 最佳实践

1. 定期审查反思报告

# 每周查看反思报告
cat memory/reflections/weekly-*.md

关注点:

  • 本周完成了什么?
  • 哪些做得好?
  • 哪些需要改进?
  • 下周计划是什么?

2. 调整自信度阈值

场景: 系统创建的技能质量不高

解决: 提高高自信阈值

{
  "thresholds": {
    "high": 0.8  // 从 0.7 提升到 0.8
  }
}

3. 优化任务优先级

场景: P0 任务过多,系统压力大

解决: 重新评估优先级,将部分 P0 降级为 P1

4. 集成外部工具

场景: 想要集成新的技能发现源

解决: 修改 learning-direction.js,添加新的搜索源


🚨 故障排查

问题 1: 循环不执行

检查:

# 检查 cron 状态
cron list

# 检查进化引擎状态
node engines/evolution-engine.js status

解决:

# 重启循环
node engines/evolution-engine.js run

问题 2: 任务队列阻塞

检查:

# 查看任务队列
cat tasks/queue.json | jq '.tasks[] | select(.status=="in_progress")'

解决:

# 手动标记为 blocked
# 编辑 tasks/queue.json,将任务状态改为 blocked

问题 3: 反思报告未生成

检查:

# 检查反思目录
ls -la memory/reflections/

# 手动生成反思
node engines/reflection.js daily

解决:

# 检查 cron 日志
# 重新设置 cron 任务
node setup-cron.js

📚 相关资源

文档

外部资源


🎉 总结

Autonomous Learning Cycle 是一个完整的自我学习循环系统,实现:

  • ✅ 自主任务执行(17 分钟循环)
  • ✅ 学习模式提取
  • ✅ 自信度评估
  • ✅ 技能自动创建
  • ✅ 每日/每周反思
  • ✅ 学习方向生成
  • ✅ 完全自主进化

适用场景: 个人能力提升、新领域学习、团队项目管理

预期效果: 30 天后能力显著提升,90 天后成为领域专家


🚀 开始你的自主进化之旅吧!

Usage Guidance
What to check before installing: 1) Review the 'skill-creator', 'setup-cron.js' and any start/init scripts to see exactly what files they create or modify. 2) Run the code in an isolated sandbox or disposable workspace (not your main OpenClaw workspace) to observe behavior. 3) Inspect any code that calls child_process/execSync or uses 'npx' — these can fetch and run remote packages. 4) Backup your ~/.jvs/.openclaw/workspace before running; don't run on a machine with sensitive environment/configs. 5) Consider raising the confidence threshold (configs/confidence-config.json) and disabling automated skill publication until you audit auto-created skill content. 6) If you need assurance, ask the author for a minimal example that only reads data (no writes or cron) and for a clear description of what new skill files will contain and where they're written.
Capability Analysis
Type: OpenClaw Skill Name: autonomous-learning-cycle Version: 1.0.0 The skill implements an autonomous 'self-evolution' framework that utilizes high-risk capabilities, including automated code generation (engines/skill-creator.js), persistence via cron job registration (setup-cron.js), and shell command execution via execSync (engines/learning-direction.js). While these features align with the stated purpose of a self-improving agent, the ability to automatically write new SKILL.md files and schedule system-level tasks creates a significant attack surface for potential prompt injection or unintended command execution. No explicit evidence of intentional malice, such as data exfiltration or hardcoded backdoors, was identified.
Capability Assessment
Purpose & Capability
The name/description (autonomous 17-minute learning loop that extracts patterns, evaluates confidence and auto-creates skills) aligns with the included engine files: evolution, extractor, confidence, reflection, learning-direction and skill-creator. However the SKILL metadata claimed to be 'instruction-only' but the package includes ~8 engine scripts that will read/write workspace files and create skills — this mismatch is noteworthy.
Instruction Scope
Runtime instructions instruct you to run init/setup-cron/start scripts and CLI engine commands. The code reads and writes many workspace paths (skills/, memory/, tasks/, instincts/, .learnings/, etc.), appends patterns/lessons, writes evolution logs, and (per learning-direction) executes shell commands (execSync 'npx skills find', and other npx/CLI usage). It also auto-adds tasks and (via skill-creator) can create skill files. These actions go beyond passive analysis and allow modification of local files and automated network-invoked tooling — so the runtime scope is broad and file-system mutating.
Install Mechanism
There is no formal install spec (no package manager download) and the README suggests copying/cloning the repo. The code itself invokes external tooling via npx (which may fetch packages from npm at runtime). No direct downloads from arbitrary URLs are present in the reviewed snippets, but npx/child_process use implies possible network fetches at runtime.
Credentials
The skill declares no required environment variables or credentials, which matches the files shown. Internally it relies on HOME or OPENCLAW_WORKSPACE to locate and modify the workspace. While it doesn't request secrets, it will access many local files (task queues, memory, skill directories). Documentation mentions ClawHub tokens for publishing, but the engines do not require credentials in the examined files.
Persistence & Privilege
always:false (good) but the skill includes setup-cron.js to register cron jobs (*/17 * * * *) and auto-create skills/files in the workspace. That gives the skill ongoing scheduled execution and the ability to change code/assets on disk (self-modifying or self-extending behavior). Autonomous invocation combined with file writes and cron registration raises the blast radius and warrants caution.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install autonomous-learning-cycle
  3. After installation, invoke the skill by name or use /autonomous-learning-cycle
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
Initial release: complete 17-minute autonomous learning/evolution system. - Implements six core engines: evolution engine, learning extractor, confidence assessment, skill creator, reflection engine, and learning direction generator. - Supports four main scheduled tasks for autonomous cycles, daily/weekly reflection, and new learning direction generation. - Provides a self-evolving workflow: autonomous task selection, execution, pattern extraction, confidence evaluation, skill generation, and reflection. - Includes setup/usage instructions, configuration options, and troubleshooting guidance. - Designed for personal growth, skill acquisition, and advanced autonomous project management.
Metadata
Slug autonomous-learning-cycle
Version 1.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is Autonomous Learning Cycle?

实现每17分钟自主执行任务、提取模式、评估自信、自动创建技能并生成每日/每周反思与新学习方向的闭环系统。 It is an AI Agent Skill for Claude Code / OpenClaw, with 133 downloads so far.

How do I install Autonomous Learning Cycle?

Run "/install autonomous-learning-cycle" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Autonomous Learning Cycle free?

Yes, Autonomous Learning Cycle is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Autonomous Learning Cycle support?

Autonomous Learning Cycle is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Autonomous Learning Cycle?

It is built and maintained by pagoda111king (@pagoda111king); the current version is v1.0.0.

💬 Comments