RangeKing Self-Evolving Agent Skill Model Card
这是可索引项目报告证据页：它保留 RangeKing Self-Evolving Agent Skill Model Card 的源材料入口、机制线索和限制提醒；正文仍需 reader/editor 与 academic public-copy review 后才能当作最终结论引用。
RangeKing Self-Evolving Agent Skill Model Card
| Field | Value |
|---|---|
| Repository | RangeKing/self-evolving-agent |
| Category | OpenClaw Self-Evolving Skill |
| Stars / forks snapshot | 9 / 0 |
| Language | Python |
| License | MIT |
| Raw capture | raw-github/rangeking_self-evolving-agent.md |
| Updated by | hourly public metadata update, 2026-05-26 05:44 +0800 |
1. Role in Self Evolve
This repository is a small but directly relevant skill-layer artifact: it turns a reactive self-improvement log into curriculum, evaluation, transfer, and promotion surfaces for OpenClaw-like agents.
2. Working Principle
agent run -> .evolution workspace -> diagnosis/evaluation agenda -> promotion or read-only legacy retention -> future hooks use promoted capabilities
3. Evidence Path
Web-observed GitHub evidence showed 9 stars, 0 forks, 18 commits, MIT license, and Python-dominant code. Visible folders include benchmarks, evals, hooks/openclaw, modules, scripts, and system, which makes the repository a concrete skill-harness bridge rather than only a README claim.
4. Teaching Use
Use this card to show the difference between unstructured “agent learned something” logs and a governed evolution workspace with migration, evaluation, and promotion.
5. Limits
The repository was not installed into OpenClaw and no benchmark scripts were run. Shell GitHub API DNS failed, so metadata is web-observed rather than API-verified.