Manager Self-Evolution

Enable managers to self-diagnose, identify defects, track improvements, and maintain skill health without relying on external prompts.

Security Scan

The skill's code and instructions are internally consistent with its stated purpose: it reads local manager conversation/memory files, checks skill integrity, and writes a local evolution log — it does not request credentials or perform network I/O.

Purpose & Capability

The skill says it will self-diagnose, find defects, track improvements, and maintain skill health. The provided self-check.py implements exactly those behaviors: it reads recent conversation logs, MEMORY.md, SOUL.md, scans the skills directory for inconsistencies, and writes entries to a local evolution-log.md. The resources accessed (workspace memory and skills folder) are appropriate for the stated purpose.

Instruction Scope

SKILL.md instructs the agent to perform read-first diagnostics and record findings to evolution-log; the script follows that scope. It reads specific files under the workspace and prints/logs findings; it does not modify MEMORY.md/SOUL.md/AGENTS.md and contains no network calls or external endpoints. The script does write a local log file and prints results when run (manual or automated trigger).

Install Mechanism

There is no install spec and no external downloads. The skill is instruction-only plus a small Python script bundled in the skill — low-risk from an install perspective.

Credentials

The skill requires no environment variables or external credentials, which is proportional. However it reads files from /root/.openclaw/workspace and enumerates /root/.openclaw/skills — these locations can contain user conversation content or other skill files (potentially sensitive). Reading those files is justified by the purpose but does grant access to potentially private data; users should be aware the script inspects workspace memory and other skill directories.

Persistence & Privilege

The skill is not always-enabled, does not claim elevated privileges, and does not modify other skills' configuration. Its only persistent effect is creating/appending a local evolution-log.md under the workspace memory directory, which is consistent with its goals.

Assessment

This skill appears to do what it claims: local self-diagnostics and a local evolution log. Before enabling it (especially with autonomous invocation): 1) Review the bundled self-check.py yourself (you already have it) to confirm no network calls or unexpected behavior (the script has none). 2) Note it reads files under /root/.openclaw/workspace and lists all skills — those locations may contain private conversations or other sensitive data, so only enable it if you're comfortable allowing a skill to read those files. 3) Consider running it manually first (python3 self-check.py diagnose) to see what it logs. 4) If you want stricter isolation, run the script in a sandbox or adjust the WORKSPACE path to a restricted directory. 5) If you expect the agent to act on discovered issues automatically, ensure user-confirmation flows are enforced (the SKILL.md promises user confirmation for major changes but the script only logs findings).

Self-Evolution Skill - 自我进化技能

功能定位

使Manager具备自我诊断、自我改进的机制，不依赖外部提醒持续优化。

核心功能

1. 自我诊断 (self-diagnose)

• 检查最近对话记录，识别理解错误模式
• 检查MEMORY.md中的教训是否被遵守
• 检查SOUL.md原则是否在行为中体现

2. 缺陷发现 (defect-discovery)

• 在heartbeat时主动扫描最近对话
• 发现问题立即记录到evolution-log.md
• 不等用户提醒

3. 改进追踪 (improvement-tracking)

• 记录每次自我发现的问题
• 追踪改进是否落实
• 评估改进效果

4. 技能健康 (skill-health)

• 定期检查skill文件完整性
• 检查skill中的承诺是否兑现
• 确保skill不成为空壳

触发机制

触发条件	执行内容
每次heartbeat	运行自我诊断
发现重大失误后	立即记录并提醒
每周五	汇总进化日志

文件结构

self-evolution/
├── SKILL.md           # 本文件
├── self-check.py      # 自我检查脚本
└── evolution-log.md   # 进化日志（自动创建）

注意事项

• 只读优先：尽可能减少写入
• 不影响现有系统：不修改MEMORY.md/SOUL.md/AGENTS.md
• 用户确认：重大改进需用户确认才执行
• 透明记录：所有自我发现都记录在evolution-log

使用方式

# 手动运行自我检查
python3 skills/self-evolution/self-check.py diagnose

# 查看进化日志
cat memory/evolution-log.md