OpenClaw自动进化系统
v1.0.0OpenClaw 自进化系统 - 让AI具备自我学习与自我提升能力 通过「犯错→学习→提炼→强化」的闭环机制，实现真正的自我改进 核心能力：自动发现问题、分析根因、提出建议、验证效果、落地形成规则
⭐ 0· 175·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
Name/description promise a self-evolution system; the provided Python implements health checks, rule-summary reads, and a learning-record writer under a workspace directory. There are no unrelated credentials, binaries, or external accesses requested.
Instruction Scope
SKILL.md advertises features like layered memory management, automatic rule‑library initialization, and cross-session persistence; the code implements a basic file-backed learning recorder, a health checker, and rule-summary reader but does not contain explicit initialization of a rules DB or advanced memory layer logic described in the docs. This is a mismatch (feature claims > implemented behavior).
Install Mechanism
No install spec and only a small Python file are included. No network downloads or package installs are performed by the skill itself, so install risk is low.
Credentials
The code uses a single optional env var OPENCLAW_WORKSPACE to locate its workspace (default /root/.openclaw/workspace). No credentials, tokens, or unrelated env vars are requested. Note: the default path under /root may be surprising; it will create/read files there unless overridden.
Persistence & Privilege
The skill reads/writes files only under its declared workspace path (STATE_DIR and .learnings). always is false and it does not modify other skills or global agent config. Persistence is local to its workspace.
Assessment
This skill appears to do what it claims at a basic level and does not request secrets or network access. Before installing: (1) decide where files will be written by setting OPENCLAW_WORKSPACE to a dedicated, non-root directory you control (the default is /root/.openclaw/workspace); (2) be aware SKILL.md promises extra features (automatic rule initialization, layered memory) that are not implemented in the shipped Python — treat it as a lightweight, file-backed helper rather than a full self‑evolution framework; (3) if you want stronger guarantees, review or run the Python in a sandbox to confirm behavior and inspect any files it creates (auto_learned_rules.md, daily learning files).Like a lobster shell, security has layers — review code before you run it.
auto-evolutionvk976sm50jx7kxbd42wzjq94fxx84ewyrautomationvk976sm50jx7kxbd42wzjq94fxx84ewyrlatestvk976sm50jx7kxbd42wzjq94fxx84ewyrlearningvk976sm50jx7kxbd42wzjq94fxx84ewyr
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
Runtime requirements
🧠 Clawdis