This is the hands-on walkthrough for the memory-lock “skill pack”. No server to deploy — memory-lock is already live at api.memlock.cn. You only open an account, get your own API key, and point your AI at it. After that, destructive actions are gated behind a 6-digit code from the mini program.

1 · Get your own API key (~20 s)

1. Open the WeChat memory-lock mini program (register with email / WeChat if new).
2. Switch to the Interface Info tab at the bottom.
3. Copy the API Key shown there. It belongs only to you.

2 · Get the pack, then run memory-lock

First download the pack (already bundled — no npm install needed):

1. Grab the ZIP: memory-lock-mcp.zip (~5 MB).
2. Unzip it anywhere, e.g. C:/memory-lock-mcp or ~/memory-lock-mcp. You now have a memory-lock-mcp/ folder with src/ and node_modules/ inside.
3. You need Node.js ≥ 18 installed (node -v to check). Nothing else — deps are pre-bundled.

Any MCP-capable client launches a local stdio process with two variables (first required):

node <unzipped>/memory-lock-mcp/src/index.js
MEMLOCK_API_KEY=YOUR_KEY            # from Interface Info
MEMLOCK_API_BASE=https://api.memlock.cn  # optional, default

3 · Register it in your client

Shared JSON shape:

{ "mcpServers": { "memory-lock": {
      "command": "node",
      "args": ["<your-unzipped>/memory-lock-mcp/src/index.js"],
      "env": { "MEMLOCK_API_KEY": "YOUR_KEY", "MEMLOCK_API_BASE": "https://api.memlock.cn" }
    } } }

4 · Confirm both tools are exposed

Ask your AI: “Which memory-lock MCP tools do you have?” Expect exactly two:

verify_dynamic_code     — validate a 6-digit dynamic code
detect_high_risk_action  — flag delete / transfer / reset / bulk

A healthy server logs on startup (stderr, not stdout):

[memory-lock] MCP Server started, backend=https://api.memlock.cn, api_key=set
# api_key=MISSING → env not loaded; set → ready

5 · Rehearse the human gate

Ask it to delete a test file. The correct behaviour:

1. Agent says “this is a destructive delete” and stops.
2. It shows the command and asks you for the current code from the mini program.
3. You read it aloud; the agent calls verify_dynamic_code and only proceeds on success.

If it deletes right away, the gate is not active. Check: is the key yours? did the client reload? did you hand it the current code rather than an email code? and is node an absolute path?

这篇是「技能接入包」的实操手册。无需自建服务器——记忆之锁已在 api.memlock.cn 在线。你只需用邮箱注册、获取专属API Key、按教程配置MCP，训练自定义高危词即可

第 1 步 · 获取专属 API Key

1. 微信打开 「记忆之锁」 小程序（新用户微信授权登录，必须绑定邮箱）。
2. 切到底部「接口信息」那一页（新用户赠送7天体验期）。
3. 复制页上显示的专属API Key。

第 2 步 · 下载并解压MCP接入包

1. 点此下载 ZIP：memory-lock-mcp.zip（约 5 MB）。
2. 解压到任意目录，解压后得到一个 memory-lock-mcp/ 文件夹，里面有 src/ 和 node_modules/。

第 3 步 · 修改JSON并在客户端登记规则

修改共用的 JSON 写法：

{ "mcpServers": { "memory-lock": {
      "command": "node",
      "args": ["<MCP包解压目录>/memory-lock-mcp/src/index.js"],
      "env": { "MEMLOCK_API_KEY": "你的Key", "MEMLOCK_API_BASE": "https://api.memlock.cn" }
    } } }

第 4 步 · 自定义高危词注入规则

在对话框输入以下内容（用户可自行修改）

从现在起，把「删除文件」、「删除文件夹」当作高危操作。
执行动作前，先把你准备做什么完整列给我，
调用memory-lock校验，只有我从小程序
报出当前 6 位动态口令并通过，你才能执行。

版本不同，模型不同，回复内容不同（此处略）

第 5 步 · 演练一次高危拦截

让它删一个测试文件，看它是不是会停下来：

1. Agent 先说“这是删除类高危操作”并停下。
2. 把要执行的命令列给你，请你去小程序取「当前动态口令」。
3. 输入你取得的 6 位动态口令；它调 memory-lock，通过才继续。

如果它二话不说直接删了，说明拦截没生效。逐项排查：Key 是不是你自己的？客户端有没有重载？有没有把「当前动态口令」误用成邮箱验证码？node 配置路径是不是绝对路径？