inner-life-memory
Transform passive logging into active development.
Requires: inner-life-core
Prerequisites Check
Before using this skill, verify that inner-life-core has been initialized:
- 1. Check that
memory/inner-state.json exists - Check that
memory/drive.json exists
If either is missing, tell the user: "inner-life-core is not initialized. Install it with clawhub install inner-life-core and run bash skills/inner-life-core/scripts/init.sh." Do not proceed without these files.
What This Solves
Without memory continuity:
CODEBLOCK0
With inner-life-memory:
CODEBLOCK1
Post-Session Flow
After each session, run this 5-step reflection:
1. Reflect
Analyze the session: what happened, what mattered, what surprised you.
2. Extract
Pull structured memories with types and confidence:
| Type | Description | Persistence |
|---|
| INLINECODE4 | Declarative knowledge | Until contradicted |
| INLINECODE5 |
Likes, dislikes, styles | Until updated |
|
relationship | Connection dynamics | Long-term |
|
principle | Learned guidelines | Stable |
|
commitment | Promises, obligations | Until fulfilled |
|
moment | Significant episodes | Permanent |
|
skill | Learned capabilities | Cumulative |
|
question | Things to explore | Until resolved |
3. Integrate
Update MEMORY.md with extracted memories. Use synapse tags for connections:
- -
<!-- updates: previous fact --> when updating - INLINECODE13 when correcting
4. Question
Generate genuine follow-up questions from the session. Not performative — real curiosity.
5. Surface
When user returns, present relevant pending questions naturally (max 3).
Confidence Scores
| Level | Range | Meaning |
|---|
| Explicit | 0.95-1.0 | User directly stated |
| Implied |
0.70-0.94 | Strong inference from context |
| Inferred | 0.40-0.69 | Pattern recognition |
| Speculative | 0.0-0.39 | Tentative, needs confirmation |
Use confidence to decide when to state facts vs ask for confirmation.
Curiosity Backlog
Maintain memory/questions.md with three sections:
CODEBLOCK2
Rules:
- - Brain Loop Step 6 adds new questions/leads
- Evening Session reviews and curates
- Dead Ends older than 30 days → archive
- Questions resolved → move to Dead Ends with result
State Integration
Reads: inner-state.json, drive.json, daily notes, diary
Writes:
- - drive.json → new seeking topics from curiosity
- inner-state.json → curiosity.recentSparks when discovering something
- questions.md → new questions, resolved dead ends
- MEMORY.md → integrated memories
When Should You Install This?
Install this skill if:
- - Your agent forgets who you are between sessions
- You want structured memory with confidence levels
- You want genuine curiosity that carries forward
- Your agent reads notes but doesn't truly remember
Part of the openclaw-inner-life bundle.
Requires: inner-life-core
inner-life-memory
将被动记录转化为主动发展。
需要：inner-life-core
前置检查
使用此技能前，请确认 inner-life-core 已初始化：
- 1. 检查 memory/inner-state.json 是否存在
- 检查 memory/drive.json 是否存在
如果缺少任一文件，请告知用户：inner-life-core 未初始化。请使用 clawhub install inner-life-core 安装，并运行 bash skills/inner-life-core/scripts/init.sh。 缺少这些文件时请勿继续。
解决的问题
没有记忆连续性时：
会话结束 → 记录笔记 → 下次会话读取笔记 → 表现出熟悉感
使用 inner-life-memory 时：
会话结束 → 运行反思 → 整合记忆 → 生成问题
下次会话 → 加载进化状态 → 呈现问题 → 真正的好奇心
会话后流程
每次会话后，执行以下 5 步反思：
1. 反思
分析本次会话：发生了什么、什么重要、什么让你惊讶。
2. 提取
提取结构化记忆，包含类型和置信度：
喜好、厌恶、风格 | 直到更新 |
| 关系 | 互动动态 | 长期 |
| 原则 | 学到的准则 | 稳定 |
| 承诺 | 承诺、义务 | 直到履行 |
| 时刻 | 重要事件 | 永久 |
| 技能 | 学到的能力 | 累积 |
| 问题 | 待探索事项 | 直到解决 |
3. 整合
将提取的记忆更新到 MEMORY.md。使用突触标签建立连接：
4. 提问
从会话中生成真正的后续问题。不是表演性的——而是真正的好奇。
5. 呈现
用户返回时，自然地呈现相关待处理问题（最多 3 个）。
置信度评分
0.70-0.94 | 从上下文强烈推断 |
| 推断 | 0.40-0.69 | 模式识别 |
| 推测 | 0.0-0.39 | 试探性，需确认 |
使用置信度决定何时陈述事实，何时请求确认。
好奇心积压
维护 memory/questions.md，包含三个部分：
markdown
开放问题
- - [问题] — 来源：[梦境/阅读/工作] — 日期
线索（半成型想法）
死胡同（不再重复）
- - [主题] — 探索于 [日期]，结果：[无/死胡同]
规则：
- - 大脑循环第 6 步添加新问题/线索
- 晚间会话审查并整理
- 超过 30 天的死胡同 → 归档
- 已解决的问题 → 移至死胡同并附上结果
状态整合
读取： inner-state.json、drive.json、每日笔记、日记
写入：
- - drive.json → 从好奇心生成新的探索主题
- inner-state.json → 发现新事物时更新 curiosity.recentSparks
- questions.md → 新问题、已解决的死胡同
- MEMORY.md → 整合的记忆
何时应安装此技能？
在以下情况下安装此技能：
- - 你的代理在会话之间忘记你是谁
- 你想要带有置信度级别的结构化记忆
- 你想要能够延续的真正好奇心
- 你的代理读取笔记但并未真正记住
属于 openclaw-inner-life 套件的一部分。
需要：inner-life-core
