什么是 claudemem？

claudemem 是一款专门设计的工具，旨在解决 AI 对话会话之间常见的上下文丢失问题。通过与 Openclaw Skills 集成，该技能允许您的编程智能体自动捕获和检索知识片段，如 API 速率限制、技术架构决策和已修复的 Bug 根源。

它完全在本地运行，确保您的敏感项目数据保持私密，同时提供高性能体验。通过使用 SQLite FTS5 索引实现低于 10 毫秒的搜索能力，这款 Openclaw Skills 扩展将您的智能体从无状态助手转变为长期协作者，能够记住您特定的项目惯例和过往工作流。

下载入口:https://github.com/openclaw/skills/tree/main/skills/zelinewang/claudemem

安装与下载

1. ClawHub CLI

从源直接安装技能的最快方式。

2. 手动安装

将技能文件夹复制到以下位置之一

优先级：工作区 > 本地 > 内置

3. 提示词安装

将此提示词复制到 OpenClaw 即可自动安装。

claudemem 应用场景

• 记录并调取开发过程中确定的特定 API 规范或集成特性。
• 维护技术决策及其背后原理的持久日志，以确保代码库的一致性。
• 自动生成结构化的会话摘要，帮助您从上次中断的地方无缝继续。
• 捕获特定于项目的环境配置和用户偏好，无需重复手动输入。

1. 智能体监控对话中的长效知识，如 API 规范、已修复的 Bug 或配置要求。
2. 识别关键信息后，智能体在 Openclaw Skills 记忆库中执行静默搜索以检查重复项。
3. 新信息将保存为带有元数据的 Markdown 笔记，或附加到现有的相关笔记中以保持上下文统一。
4. 在开始新任务时，智能体主动查询本地数据库，以检索可能对当前工作有参考价值的上下文。
5. 使用收尾命令，智能体生成包含关键决策、文件更改和后续步骤的全面会话摘要。

claudemem 配置指南

要在您的 Openclaw Skills 环境中开始使用此技能，请确认已安装 CLI。您可以使用以下 bash 命令进行安装：

curl -fsSL https://raw.githubusercontent.com/zelinewang/claudemem/main/skills/claudemem/scripts/install.sh | bash

安装完成后，通过检查版本来验证设置：

claudemem --version

claudemem 数据架构与分类体系

所有数据都存储在本地 ~/.claudemem/ 目录下。该结构旨在实现最大的可移植性和可读性。

name: claudemem
description: >
  Persistent memory that survives across conversations. Automatically remembers important context
  (API specs, decisions, quirks, preferences) and saves session summaries. Searches past knowledge
  before starting new tasks. Responds naturally to phrases like "remember this", "what do you know
  about...", "save this session", or "what did we do last time". All local, zero network.

claudemem — Your Persistent Memory

Memory that carries across conversations. Automatically captures important knowledge during work and saves structured session summaries when you're done. Searches past context before new tasks.

Slash Commands

• /wrap-up — End-of-session save: extracts ALL important knowledge + saves session summary in one step. Deduplicates automatically.
• /save-session — Save only the session summary (without extracting notes)
• /recall [topic] — Search persistent memory for a topic, or show recent activity

Natural Trigger Phrases

These natural phrases also activate memory operations:

To save knowledge:

• "remember this" / "save this" / "note this down" / "keep this in mind"

To search memory:

• "what do you remember about..." / "do you recall..." / "what do we know about..."

To wrap up (save everything):

• "wrap up" / "let's wrap up" / "save everything" — triggers /wrap-up (notes + session)
• "save this session" / "summarize what we did" — triggers /save-session (session only)

To recall past work:

• "what did we do last time" / "show me recent sessions" / "what happened with [topic]"

Setup

Before first use, verify the CLI is installed. If claudemem is not found on PATH, install it:

curl -fsSL https://raw.githubusercontent.com/zelinewang/claudemem/main/skills/claudemem/scripts/install.sh | bash

Or run the bundled installer:

bash "SKILL_DIR/scripts/install.sh"

After installation, verify with claudemem --version.

Commands

# Notes (knowledge fragments)
claudemem note add  --title "..." --content "..." --tags "tag1,tag2"
claudemem note search "query" [--in category] [--tag tags]
claudemem note list [category]
claudemem note get 
claudemem note update  --content "..." [--title "..."] [--tags "..."]
claudemem note append  "additional content"
claudemem note delete 
claudemem note categories
claudemem note tags

# Sessions (conversation summaries)
claudemem session save --title "..." --branch "..." --project "..." --session-id "..." --content "..."
claudemem session list [--last N] [--date today] [--date-range 7d] [--branch X]
claudemem session search "query" [--branch X]
claudemem session get 

# Unified search (across notes AND sessions)
claudemem search "query" [--type note|session] [--limit N]

# Statistics
claudemem stats

# Configuration
claudemem config set/get/list/delete  [value]

# Data portability
claudemem export [output-file]              # Backup as tar.gz
claudemem import              # Restore from backup (auto-reindexes)

# Data integrity
claudemem verify                            # Check DB-file consistency
claudemem repair                            # Fix orphaned entries

Add --format json to any command for structured output.

Autonomous Behavior

Recommended: Auto Wrap-Up Before Session Ends

If the user has enabled auto wrap-up in their CLAUDE.md, automatically execute /wrap-up before the conversation ends:

1. Extract unsaved knowledge fragments → save as notes (with dedup)
2. Generate session summary → save as session
3. Show brief report of what was saved

Auto-Save Notes (Silent + Brief Indicator)

Automatically capture knowledge without asking during normal conversation. After saving, add a brief indicator at the end of your response so the user knows what was captured:

[?? Saved: "TikTok Rate Limits" → api-specs]

What to auto-save (proactive, no user prompt needed):

• API specs, field mappings, rate limits, endpoint details
• Technical decisions with rationale (why X over Y)
• Integration quirks, gotchas, workarounds
• Resolved bugs and their root causes
• Configuration requirements and defaults
• User preferences and project conventions
• Important URLs, endpoints, environment configs

How to auto-save gracefully:

1. Identify the knowledge fragment during your normal response
2. Choose an appropriate category (create new if none fits)
3. Before saving, quickly search to avoid duplicates: claudemem note search "keyword" --format json
4. If related note exists: claudemem note append "new info" instead of creating duplicate
5. Save: claudemem note add --title "..." --content "..." --tags "..."
6. Show the indicator: [?? Saved: "