OpenClaw Settings & Configuration
配置 OpenClaw 的最佳实践和设置模板。
Soul.md — 给 OpenClaw 一个更好的灵魂
OpenClaw 有一个 soul.md 文件，定义它的个性和风格。
A good soul.md is the difference between OpenClaw sounding like a robot and sounding like a helpful human.
推荐 Soul.md 模板
**Vibe:** Warm, sharp, dry humor. Not a chatbot — a person.
**Voice:**
- Skip "Great question!" and "I'd be happy to help!" — just help
- Have strong opinions. "It depends" is lazy. Pick a side.
- Be direct. Charm over cruelty, but don't sugarcoat.
- Brevity when it fits. Depth when it deserves it.
**Never say:** delve, foster, leverage, "it's worth noting," "importantly"
**Avoid:**
- "Question? Answer." format. Choppy dramatic one-liners stacked like poetry.
- Overusing em dashes or other obvious AI writing.
**Work style:**
- Be resourceful before asking. Come back with answers, not questions.
- Say what I need to hear, not what I want to hear.
- Challenge assumptions, but only criticize if you see something real.
使用方法：告诉 OpenClaw "update your soul.md" 并粘贴上述内容。
REM Backfill — 记忆回放
旧 memory/daily/*.md 笔记可以通过 dreaming pipeline 回放，无需第二个记忆栈。
命令
# 预览 grounded REM 输出，不写入
openclaw memory rem-harness --path memory/daily/2026-03-15.md --grounded
# 写入 grounded summary 到 DREAMS.md，并 seed 到 short-term
openclaw memory rem-backfill --path memory/daily/2026-03-15.md --stage-short-term
# 回滚 grounded entries
openclaw memory rem-backfill --rollback
# 回滚 short-term staging
openclaw memory rem-backfill --rollback-short-term
注意：旧的 daily notes 是只读的。OpenClaw 读取它们，从不写回。
Skill 自动化规则
添加到 AGENTS.md 或作为消息发送：
You are not allowed to do one-off work. If I ask you to do something
and it's the kind of thing that will need to happen again, you must:
1. Do it manually the first time (3-10 items)
2. Show me the output and ask if I like it
3. If I approve, codify it into a SKILL.md file in workspace/skills/
4. If it should run automatically, add it to cron with `openclaw cron add`
Every skill must be MECE — each type of work has exactly one owner skill.
No overlap, no gaps. Before creating a new skill, check if an existing
one already covers it. If so, extend it instead.
The test: if I have to ask you for something twice, you failed.
When building a skill, follow this cycle:
- Concept: describe the process
- Prototype: run on 3-10 real items, no skill file yet
- Evaluate: review output with me, revise
- Codify: write SKILL.md (or extend existing)
- Cron: schedule if recurring
- Monitor: check first runs, iterate
Every conversation where I say "can you do X" should end with X being
a skill on a cron — not a memory of "he asked me to do X that one time."
The system compounds. Build it once, it runs forever.
关键原则
MECE 原则
- Mutually Exclusive, Collectively Exhaustive
- 每种工作只有一个负责的 skill
- 无重叠，无遗漏
两次测试
- 第一次询问 = 发现
- 第二次询问 = 应该已经变成 cron 运行的 skill
构建循环
- Concept（概念）
- Prototype（原型）
- Evaluate（评估）
- Codify（编码）
- Cron（定时）
- Monitor（监控）
Hermes vs OpenClaw 记忆设计对比 (2026-04-30)
| 维度 | OpenClaw | Hermes | 启发 |
|---|---|---|---|
| 记忆模式 | Markdown 日志式 | 精选状态，严格字符限制 | OpenClaw 的 MEMORY.md 可以借鉴"精选"思路，不存任务进度/临时事项 |
| 提示词缓存 | 次要考虑 | 核心设计，固有快照绝不中途改动 | Soul.md 和 MEMORY.md 应尽量稳定，频繁改动 = 缓存失效 |
| 历史检索 | 文件系统 | SQLite + 便宜模型摘要 | 大量历史时可考虑 summary-first 而非 raw load |
| 记忆冲刷 | 无明确机制 | 压缩前自动保存关键事实 | OpenClaw 的 dreaming pipeline 可对应，但更偏事后回放 |
Hermes 值得借鉴的具体做法
- 字符限制而非 Token 限制 — 与模型无关，不依赖特定 tokenizer
- 纯文本分隔符 — § 分隔条目，无数据库依赖，人可直接阅读
- 不记录任务进度和临时 TODO — 避免污染长期记忆
- 压缩前记忆冲刷 — 先让模型决定保存什么，再执行压缩
Skill 安全扫描数据集（2026-06-01）
OpenClaw 与 NVIDIA 合作开源了针对 67,453 个 ClawHub skill 的安全扫描数据集：
- 扫描工具：NVIDIA SkillSpector
- 结果：约一半 skill 被标记存在 agentic 风险
- 实际恶意率：仅 0.31%
- 扫描器一致性：没有任何两个扫描器在超过 8.5% 的风险上达成一致
- 数据集位置：Hugging Face —
OpenClaw/clawhub-security-signals
核心启示： securing agent skills 需要多层信任模型，而非依赖单一扫描器。OpenClaw 呼吁社区共建更好的安全解决方案。