← 返回 Skills 市场
Mem Skill
作者
Oliver Guo
· GitHub ↗
· v1.2.0
· MIT-0
175
总下载
0
收藏
0
当前安装
2
版本数
在 OpenClaw 中安装
/install mem-skill
功能描述
Self-evolving memory and knowledge accumulation system for AI agents. Acts as a persistent 'second brain' that automatically retrieves past experiences, capt...
安全使用建议
This skill appears to do what it says: store and retrieve local knowledge and optionally use QMD for semantic search. Before installing or running:
- Back up your workspace (init/upgrade will read/modify many markdown files and may insert fields).
- If you choose the QMD engine, be prepared for a prompted global npm install and large model downloads; review @tobilu/qmd before installing.
- Note the skill can read and write everything under the workspace directory and will index files you allow; avoid initializing it in folders containing secrets.
- Prefer project-scoped QMD collections (not global) if you want to keep collections isolated per-project.
- If you need stronger guarantees, inspect the included init.sh and README yourself, and run init/interactions manually rather than allowing unattended/autonomous installs or upgrades.
功能分析
Type: OpenClaw Skill
Name: mem-skill
Version: 1.2.0
The mem-skill bundle is a legitimate persistent memory and knowledge management system for AI agents. It facilitates the creation and retrieval of local Markdown-based knowledge bases and experience logs. While the skill performs system-level operations such as file I/O, shell command execution (e.g., `qmd`, `npm`), and optional package installation in `scripts/init.sh` and `SKILL.md`, these actions are transparently documented and triggered by explicit user commands (/mem-skill init). The logic includes safeguards such as mandatory user confirmation before recording data and clear logging of operations in `log.md`. No evidence of data exfiltration, malicious prompt injection, or unauthorized persistence was found.
能力评估
Purpose & Capability
Name/description match the behavior: it builds a local knowledge/experience store, indexes and searches it, and optionally integrates with QMD. Required actions (creating knowledge-base/, writing .mem-skill.config.json, running qmd commands) are proportionate to a memory engine.
Instruction Scope
SKILL.md and the init/upgrade script instruct the agent to create files and scan the workspace's knowledge-base/ and experience/ MD files, backfill fields, and ingest external sources/URLs when requested. This is coherent for a memory system but means the skill will read and modify workspace files and can fetch external URLs or run qmd commands that index many files — users should expect broad read/write access within the workspace and interactive prompts before global actions.
Install Mechanism
There is no automated install spec in the registry, but the repo includes an init.sh that may invoke 'npm install -g @tobilu/qmd' (prompted interactively) and other qmd commands. Installing QMD pulls code from npm and may download GGUF models (~GBs). This is expected for the optional QMD engine but carries the usual risks of running third-party packages and large model downloads.
Credentials
The skill declares no environment variables or credentials and does not request unrelated secrets. It uses standard local tools (bash, sed, python3, npm, qmd) which are reasonable for its functionality. No evidence it tries to access unrelated credentials or system-wide config.
Persistence & Privilege
always is false and the skill is user-invocable; it can be invoked autonomously by the agent (platform default). It writes only into the workspace (knowledge-base/, experience/, log.md, .mem-skill.config.json) and does not attempt to modify other skills or global agent settings beyond optionally installing qmd globally (npm -g) when the user consents.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install mem-skill - 安装完成后，直接呼叫该 Skill 的名称或使用
/mem-skill触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.2.0
Version 1.2.0 inspired by LLM Wiki (Andrej Karpathy)
- Added logging support: new log.md file records all initialization and upgrade operations.
- Introduced /mem-skill upgrade command for safe, idempotent upgrades and metadata migration.
- Upgrades backfill missing **Source:** and **Related:** fields to all existing entries.
- All file creations and upgrades are logged for traceability.
- Updated initialization to include log.md and to log engine selection (default or QMD).
v0.1.0
mem-skill v0.1.0 — Initial Release
- Introduces a self-evolving memory and knowledge accumulation system for AI agents, acting as a persistent "second brain."
- Provides easy initialization with both a default engine (JSON/Markdown-based) and optional QMD semantic search support.
- Adds `/mem-skill init` command with interactive or flag-driven setup for collection scopes and names.
- Supports immediate manual recording with `/mem-skill recordnow` for capturing completed tasks and experiences.
- Implements an always-on core loop to extract keywords, detect topic switches, and proactively load/use relevant experience across skills.
- Ensures all user-facing setup (especially for QMD) confirms critical details interactively—no silent assumptions or automatic collection creation.
元数据
常见问题
Mem Skill 是什么？
Self-evolving memory and knowledge accumulation system for AI agents. Acts as a persistent 'second brain' that automatically retrieves past experiences, capt... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 175 次。
如何安装 Mem Skill？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install mem-skill」即可一键安装，无需额外配置。
Mem Skill 是免费的吗？
是的，Mem Skill 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Mem Skill 支持哪些平台？
Mem Skill 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 Mem Skill？
由 Oliver Guo（@oliguo）开发并维护，当前版本 v1.2.0。
推荐 Skills