← Back to Skills Marketplace
252
Downloads
0
Stars
1
Active Installs
1
Versions
Install in OpenClaw
/install auto-reflection
Description
自动执行深度反思和学习优化。每 5 轮自动深度反思，能力评估自动更新，学习计划自动调整，知识图谱自动优化。 Triggers: reflection, 反思，auto-reflection, 能力评估，learning optimization, 学习优化
README (SKILL.md)
💭 Auto Reflection - 自动反思技能
版本: 1.0 创建时间: 2026-03-19 状态: ✅ 激活
🎯 功能说明
自动执行深度反思和学习优化：
- 每 5 轮自动深度反思
- 能力评估自动更新
- 学习计划自动调整
- 知识图谱自动优化
- HEARTBEAT.md 自动同步
🔄 触发条件
自动触发
- 每完成 5 轮学习自动触发深度反思
- 每轮学习后更新能力评估
手动触发
# 运行反思
node skills/auto-reflection/index.js
📊 反思内容
能力分析
- 整体等级和分数变化
- 各维度能力提升排序
- 优势与弱点识别
知识分析
- 知识节点增长
- 领域掌握度
- 主题完成情况
元认知
- 学习效率评估
- 知识增长率
- 改进建议生成
📝 输出产物
反思报告
D:\OpenClaw\workspace\reflection-round-N-YYYY-MM-DD.json
包含：
- 能力变化详情
- 知识图谱分析
- 成就与不足
- 下一步计划
- 元认知建议
HEARTBEAT.md 更新
自动插入最新反思内容到 HEARTBEAT.md
能力评估更新
更新 capabilities.json 的时间戳和版本号
🎯 反思维度
| 维度 | 说明 | 权重 |
|---|---|---|
| 能力提升 | 各维度分数变化 | 30% |
| 知识增长 | 知识节点和领域掌握 | 25% |
| 学习效率 | 单位轮次提升幅度 | 20% |
| 成果质量 | 实际产出和成就 | 15% |
| 改进空间 | 弱点识别和改进计划 | 10% |
📈 反思质量指标
| 等级 | 标准 | 说明 |
|---|---|---|
| S | 深度洞察 + 可执行建议 | 产生重大改进 |
| A | 清晰分析 + 具体建议 | 产生明显改进 |
| B | 基本分析 + 一般建议 | 产生小幅改进 |
| C | 表面分析 + 模糊建议 | 改进有限 |
🔧 配置参数
{
"autoReflection": {
"enabled": true,
"reflectionInterval": 5,
"updateCapabilities": true,
"updateHeartbeat": true,
"saveReport": true,
"minQualityScore": 0.7
}
}
📊 历史反思记录
| 轮次 | 日期 | 等级 | 关键发现 |
|---|---|---|---|
| 5 | - | - | - |
| 10 | - | - | - |
| 15 | - | - | - |
| ... | - | - | - |
| 45 | - | - | - |
🚀 使用示例
完整流程
学习完成 → 更新能力评估 → 检查反思条件 →
→ (满足条件) 生成反思报告 → 更新 HEARTBEAT →
→ (不满足) 跳过等待下次
反思后的行动
- 阅读反思报告
- 理解改进建议
- 调整学习计划
- 执行改进行动
- 验证改进效果
💡 最佳实践
- 定期反思: 每 5 轮不要间断
- 深度分析: 不仅看分数，要理解原因
- 可执行建议: 建议要具体可操作
- 跟踪改进: 下次反思验证改进效果
- 知识关联: 建立跨领域的知识连接
⚠️ 注意事项
- ❌ 避免形式化反思（为反思而反思）
- ❌ 避免过于笼统的建议
- ❌ 避免只分析不行动
- ✅ 保持客观和诚实
- ✅ 关注可执行的改进
- ✅ 建立持续改进循环
自动反思技能结束
Usage Guidance
This skill appears to do what it says: generate local reflection reports, update capabilities.json, and insert a summary into HEARTBEAT.md. Before installing or running: 1) Review and back up D:\OpenClaw\workspace (or change the hardcoded CONFIG paths) because the script will read/write and may overwrite files. 2) Note the code uses Windows absolute paths and synchronous file writes — it may fail or behave unexpectedly on non-Windows systems or if directories don't exist. 3) There is a clear coding bug/typo in updateHeartbeatWithReflection (nonstandard/invalid variable usage) that will likely cause the script to crash; inspect/fix the function before use. 4) Run the script first in a sandbox or test workspace to confirm behavior. 5) No secrets or network calls were found, so there is no obvious exfiltration risk, but you should still verify the code if you plan to grant the agent autonomous runs.
Capability Assessment
Purpose & Capability
The skill declares automatic reflection and learning-optimization features and the code reads/writes local workspace artifacts (capabilities.json, knowledge-graph.json, HEARTBEAT.md, reflection JSON files, logs). These file operations align with the described purpose; no unrelated services or credentials are requested.
Instruction Scope
SKILL.md instructs running the bundled Node script. The runtime instructions and code operate only on local files under a workspace path and produce reports and updates to HEARTBEAT.md and capabilities.json. There are no external network calls or attempts to read unrelated system config. Note: the SKILL.md and code both assume a specific workspace path (D:\OpenClaw\workspace) so the skill will attempt to read/write those local files.
Install Mechanism
No install spec; this is instruction-only with an included index.js. Nothing is downloaded or installed from remote sources.
Credentials
The skill requests no environment variables or credentials (proportional). However, it hardcodes Windows-specific absolute paths (D:\OpenClaw\workspace, etc.) rather than using configurable env vars — this is a design/portability issue rather than a credential overreach.
Persistence & Privilege
always is false and the skill does not modify other skills or global agent configuration. It only modifies files in its declared workspace; no persistent platform-level privileges are requested.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install auto-reflection - After installation, invoke the skill by name or use
/auto-reflection - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
Initial release: Auto reflection every 5 rounds with capability assessment
Metadata
Frequently Asked Questions
What is Auto Reflection?
自动执行深度反思和学习优化。每 5 轮自动深度反思，能力评估自动更新，学习计划自动调整，知识图谱自动优化。 Triggers: reflection, 反思，auto-reflection, 能力评估，learning optimization, 学习优化. It is an AI Agent Skill for Claude Code / OpenClaw, with 252 downloads so far.
How do I install Auto Reflection?
Run "/install auto-reflection" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Auto Reflection free?
Yes, Auto Reflection is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Auto Reflection support?
Auto Reflection is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Auto Reflection?
It is built and maintained by wyblhl (@wyblhl); the current version is v1.0.0.
More Skills