← Back to Skills Marketplace
Agent Memory Keeper
by
li-jin-xuan
· GitHub ↗
· v1.0.1
· MIT-0
78
Downloads
0
Stars
0
Active Installs
2
Versions
Install in OpenClaw
/install agent-memory-keeper
Description
AI Agent记忆修复系统。解决每次会话失忆问题——自动复盘对话、提炼关键信息存入长期记忆、下次会话自动加载。 适用于所有OpenClaw Agent，让Agent记住用户偏好、项目进展、技术决策、重要信息。 触发词：记忆、忘了、之前说过、记录一下、帮我记、你还记得
README (SKILL.md)
Agent Memory Keeper · AI Agent 记忆修复系统
为什么需要这个？ 每次新会话AI都是一张白纸。你以为它"应该记得"的事情，它全忘了。 这个Skill让Agent学会——写下来才算数。
核心功能
| 功能 | 说明 |
|---|---|
| 自动复盘 | 对话结束时自动分析 → 提炼关键信息 → 写入长期记忆 |
| 记忆加载 | 新会话启动时自动加载最近记忆 + 检查更新 |
| 知识检索 | 随时搜索历史记忆，不再说"我不记得了" |
| 记忆归档 | 定期清理过期记忆，保持知识库精炼 |
快速开始
安装
cp -r agent-memory-keeper ~/.openclaw/workspace/skills/agent-memory-keeper
初始化记忆结构
告诉你的Agent：
帮我初始化记忆系统，创建 memory/ 目录和基础结构
日常使用
对话结束时，Agent会自动执行复盘。你也可以手动触发：
复盘一下这次对话，把重要的记下来
帮我查一下之前关于XXX的记录
把这件事记下来：...
工作流程
1. 对话复盘（自动）
每次对话结束时执行：
## 复盘清单
1. 有没有新的决策/方向变化？ → 记录到MEMORY.md
2. 有没有新项目/术语？ → 写进对应文件
3. 用户有没有给反馈/批评？ → 提炼规则，更新规则库
4. 有哪些信息可以跨领域推导？ → 做知识裂变
5. 有没有用户偏好/习惯？ → 记录到USER.md
2. 记忆结构
workspace/
├── MEMORY.md # 长期记忆（手动维护的精华）
├── USER.md # 用户信息
├── memory/
│ ├── YYYY-MM-DD.md # 每日原始日志
│ ├── heartbeat-state.json
│ └── knowledge/ # 知识节点
│ ├── rules/
│ │ └── rules.md # 行为规则库
│ └── *.md # 领域知识
└── TOOLS.md # 工具笔记
3. 记忆检索
需要回忆时：
搜索记忆：关键词
# 或
查一下之前关于[主题]的记录
4. 记忆维护
定期执行：
检查记忆状态 → 清理过期信息 → 合并重复条目 → 提炼精华到MEMORY.md
最佳实践
✅ 应该记住的
- 用户的名字、称呼、偏好
- 项目的状态、决策、技术选型
- 错误教训和修复方案
- 重要的时间和日期
- 用户的批评和反馈
❌ 不需要记的
- 日常问候
- 临时性、一次性对话
- 已经完成的任务细节（记结论就行）
适用场景
- 开发者：让Agent记住项目架构、技术决策、bug修复方案
- 内容创作者：记住内容风格、受众偏好、已发布的主题
- 团队协作：多Agent共享知识库，信息不掉队
- 个人助理：记住日程、偏好、重要信息
参考文档
- 记忆规则:
references/memory-rules.md - 复盘模板:
references/review-template.md
Usage Guidance
Install only if you are comfortable with the agent writing long-term local notes about you and your projects. Before use, set explicit rules not to store secrets, credentials, health, financial, legal, or other sensitive personal information, and periodically inspect and delete MEMORY.md, USER.md, and the memory directory as needed.
Capability Assessment
Purpose & Capability
The advertised purpose is coherent: it creates and searches long-term local memory files for user preferences, project decisions, feedback, and knowledge. That capability is expected for a memory skill, but it inherently handles private conversation-derived data.
Instruction Scope
The skill uses broad trigger phrases and describes automatic end-of-conversation review and next-session loading, while giving no per-write confirmation or clear boundary for what should never be stored.
Install Mechanism
Installation is a manual copy into the OpenClaw skills directory. The included scripts are small local shell helpers with no package installation, network access, credential access, or obfuscated behavior found.
Credentials
The file access is local and mostly scoped to MEMORY.md, USER.md, and the memory directory, which fits the stated purpose. The proportionality concern is privacy retention, not system compromise.
Persistence & Privilege
The artifacts explicitly instruct persistent storage of user names, preferences, boundaries, criticism, feedback, project decisions, and daily logs, but do not provide retention, deletion, inspection, or sensitive-data exclusion guidance.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install agent-memory-keeper - After installation, invoke the skill by name or use
/agent-memory-keeper - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.1
- 修正了复盘流程示例，将“老李有没有给反馈/批评”修改为更通用的“用户有没有给反馈/批评”
- 移除了“配置说明: references/configuration.md”参考文档入口
v1.0.0
Initial release of Agent Memory Keeper, an AI memory enhancement system for OpenClaw Agents.
- Automatically reviews and extracts key information at the end of each conversation for long-term memory.
- Loads relevant memories at session start and enables seamless recall of past preferences, decisions, and important info.
- Supports manual memory capture and retrieval via trigger phrases (e.g., 记忆, 记录一下, 你还记得).
- Periodically cleans and consolidates memory data to keep the knowledge base focused and up to date.
- Works for developer, creator, team, and personal assistant scenarios.
Metadata
Frequently Asked Questions
What is Agent Memory Keeper?
AI Agent记忆修复系统。解决每次会话失忆问题——自动复盘对话、提炼关键信息存入长期记忆、下次会话自动加载。 适用于所有OpenClaw Agent，让Agent记住用户偏好、项目进展、技术决策、重要信息。 触发词：记忆、忘了、之前说过、记录一下、帮我记、你还记得. It is an AI Agent Skill for Claude Code / OpenClaw, with 78 downloads so far.
How do I install Agent Memory Keeper?
Run "/install agent-memory-keeper" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Agent Memory Keeper free?
Yes, Agent Memory Keeper is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Agent Memory Keeper support?
Agent Memory Keeper is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Agent Memory Keeper?
It is built and maintained by li-jin-xuan (@li-jin-xuan); the current version is v1.0.1.
More Skills