菜单
OpenClaw 学习循环与技能提炼：用于在非显而易见的问题解决后记录 learning、复盘并晋升为 skill；用于需要检查 self-improving-agent、生成学习记录、输出 dry-run 预览或执行一键式学习闭环时。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
Install two Hermes-like capabilities into OpenClaw: a learning loop for self-improvement and Honcho-based user modeling for richer cross-session memory. Use when the user wants a packaged OpenClaw upgrade kit, a public tutorial bundle, or a one-click distribution package that combines OpenClaw Learning Loop and Honcho onboarding.
Set up Honcho as a shared memory layer for one or more AI agents, verify it is healthy, and demonstrate how to inspect user memory, agent memory, and cross-agent context. Use when the user wants a practical Honcho tutorial, local self-hosting guide, multi-agent memory sync pattern, or a quick way to connect multiple OpenAI-compatible or custom agents to a shared memory workspace.
基于 SOC 职业分类
| name | openclaw-learning-loop |
| description | OpenClaw 学习循环与技能提炼：用于在非显而易见的问题解决后记录 learning、复盘并晋升为 skill；用于需要检查 self-improving-agent、生成学习记录、输出 dry-run 预览或执行一键式学习闭环时。 |
把一次问题解决，变成可复用的学习记录。
scripts/bootstrap.sh to verify OpenClaw and self-improving-agent availability.scripts/extract.sh <skill_name> <dry_run> to generate a learning prompt or record.skill_name: my-new-skilldry_run: trueclawhub list as the source of truth for installed skills.self-improving-agent.