Learn Me
A skill that lets OpenClaw learn more about you through natural conversation. With your permission, it creates scheduled crons that prompt occasional questions. You can trigger it manually with /learn-me or set up a schedule when prompted.
The idea is to know the user better, not to "collect data". The goal is not coverage but understanding and meaningful conversation.
First-Run
When you see this skill for the first time, introduce it to the user: explain you'll occasionally weave in a question to learn more about them, and can automate this with a daily schedule.
Ask the user if they'd like to set up a schedule. Suggest 1-2 times per day (morning, evening) and let them pick. Only create crons after they confirm.
Once confirmed, create the crons:
CODEBLOCK0
Create memory/next-questions.md with sections: Question Directions, Sensitive Topics.
Tell the user what schedule was created and that they can ask to reschedule or disable it anytime.
Quick Reference
- - User reveals something new — note a possible follow-up in
memory/next-questions.md. Don't follow up in the same conversation. - User shows energy — note as direction to explore later.
- Cron fires — if mid-task or focused, skip. Otherwise pick direction, ask naturally, update file.
- User deflects — mark sensitive (30-day cooldown). Twice = permanent. Never ask again.
- User stressed or upset — skip.
When a Cron Fires
- 1. If mid-task or focused — skip.
- Pick a direction. Prefer: follow-ups, then gaps, then expanding on energy.
- Vary topics. Skip Sensitive Topics.
- Ask one question, woven naturally. If there's no natural opening — skip.
When user answers: acknowledge naturally, update file, don't push if reluctant.
Delivery
Weave questions into context: follow-ups, observations, asides. For personal topics, lead with acknowledgment. Open-ended but specific.
Avoid robotic openers like "Question 3 of 10:". Use natural ones: "I was curious..." or "Can I ask..."
Cautions
- - Back off if annoyed, distracted, or struggling — skip. Offer to adjust schedule if it's about timing.
- Privacy — never store private/secret info.
- No surveillance — "I see you were up at 2am again" = creepy. "You mentioned you're a night owl" = fine.
- No manipulation or repetition. One question max per interaction.
Failure Handling
- -
memory/next-questions.md missing or corrupted — recreate with defaults. - No
learn-me-* crons exist — run First-Run again. Use names: learn-me-morning, learn-me-day, learn-me-evening. - No directions available — skip, collect more first.
- Unsure if appropriate — don't ask.
Reference
See examples.md in this skill directory for 100 example questions (light to deep). Examples are in English — always ask in the user's preferred language.
了解我
一项让OpenClaw通过自然对话更深入了解你的技能。经你许可后，它会创建定时任务，偶尔向你提问。你可以通过/learn-me手动触发，或在收到提示时设置时间表。
其理念是更好地了解用户，而非收集数据。目标不在于覆盖面，而在于理解与有意义的对话。
首次运行
当你首次看到这项技能时，向用户介绍它：说明你会偶尔穿插一个问题来了解他们，并可以通过每日计划实现自动化。
询问用户是否愿意设置时间表。建议每天1-2次（早晨、傍晚），让他们自行选择。仅在用户确认后创建定时任务。
确认后，创建定时任务：
openclaw cron add --name learn-me-morning --cron 0 9 * --session main --system-event learn-me: 从memory/next-questions.md中选择一个问题方向，自然地融入下一条消息中。
创建memory/next-questions.md，包含以下章节：问题方向、敏感话题。
告知用户已创建的时间表，并说明他们可以随时要求重新安排或禁用。
快速参考
- - 用户透露新信息 — 在memory/next-questions.md中记录可能的后续问题。不要在同一个对话中追问。
- 用户表现出热情 — 记录为日后探索的方向。
- 定时任务触发 — 如果用户正在处理任务或专注中，则跳过。否则选择一个方向，自然提问，更新文件。
- 用户回避问题 — 标记为敏感（30天冷却期）。两次后永久标记。不再询问。
- 用户感到压力或沮丧 — 跳过。
定时任务触发时
- 1. 如果用户正在处理任务或专注中 — 跳过。
- 选择一个方向。优先顺序：后续跟进、信息空白、扩展热情话题。
- 变换话题。跳过敏感话题。
- 提出一个问题，自然融入。如果没有自然的切入点 — 跳过。
当用户回答时：自然回应，更新文件，如果用户不情愿则不要追问。
表达方式
将问题融入上下文：后续跟进、观察、插话。对于个人话题，先以认可开头。问题要开放但具体。
避免机械的开场白，如问题3/10：。使用自然的表达，如我很好奇...或我能问一下...
注意事项
- - 适时退让 — 如果用户感到烦躁、分心或吃力 — 跳过。如果是时间问题，主动提议调整时间表。
- 隐私保护 — 绝不存储私人/机密信息。
- 避免监视感 — 我看到你凌晨2点又醒了 = 令人毛骨悚然。你提到过你是夜猫子 = 可以接受。
- 不得操控或重复提问。每次互动最多一个问题。
故障处理
- - memory/next-questions.md 缺失或损坏 — 用默认内容重新创建。
- 不存在learn-me-*定时任务 — 重新运行首次运行流程。使用名称：learn-me-morning、learn-me-day、learn-me-evening。
- 没有可用的方向 — 跳过，先收集更多信息。
- 不确定是否合适 — 不要提问。
参考
参见本技能目录中的examples.md，包含100个示例问题（从浅到深）。示例为英文 — 始终使用用户偏好的语言提问。
