Google Keep Skill — Google Keep 技能

Google Keep 技能

技能 to interact with Google Keep via nodriver (real Chrome, no 机器人 检测ion).

安装ation

Prerequisites

Python 3.11+ uv (package and 环境 管理器) Google Chrome 安装ed on the 系统 (e.g., sudo apt 安装 google-chrome-stable on Linux)

技能 Location

The 技能 must be in /path/to/google-keep-技能/ (or in the Nano机器人 workspace/技能s). Nano机器人 discovers 技能s that have 技能.md and _meta.json in this tree.

安装 Dependencies

In the 技能 root, uv uses pyproject.toml; there is no need to 运行 anything other than uv 运行 in the commands below. On the first 运行, uv 创建s the 环境 and 安装s dependencies.

cd /path/to/google-keep-技能 uv 运行 python scripts/keep.py 检查 # example; on first 运行 uv 安装s deps

Configuration in Nano机器人

Nothing in Nano机器人's config.json — there is no need to register the 技能 in a configuration file; it is used via bash commands when the user 请求s actions in Google Keep.

记录in once — before the 机器人 can 创建/列出/edit notes, it is necessary to manually 记录 in to Chrome (the 会话 is saved). The user or the 代理 must 执行:

cd /path/to/google-keep-技能 && uv 运行 python scripts/keep.py 记录in

Chrome opens; 记录 in to your Google account, close the browser. The 会话 is saved 安全ly outside the 技能 directory at ~/.config/google-keep-技能/ with restricted 权限s and reused in future calls.

How the 机器人 uses the 技能 — the 代理 calls the 运行_command (terminal) 工具 with the complete command, for example:

cd /path/to/google-keep-技能 && uv 运行 python scripts/keep.py 列出 --limit 5

Or to 创建 a note: ... keep.py 创建 --title "Title" --content "Text".

Initial 设置up (记录in — once)

执行 once to save the 会话.

cd /path/to/google-keep-技能 uv 运行 python scripts/keep.py 记录in

Chrome will open with the Google Keep page. 记录 in normally. After 检测ing the 记录in, the browser closes and the 会话 is saved.

验证 会话 cd /path/to/google-keep-技能 && uv 运行 python scripts/keep.py 检查

Clear 会话 cd /path/to/google-keep-技能 && uv 运行 python scripts/keep.py 记录out

Only use 记录out if you want to unlink the account. After this, you will need to 记录 in agAIn.

Commands

All 执行d via terminal (运行_command in MCP/nano机器人 上下文):

cd /path/to/google-keep-技能 && uv 运行 python scripts/keep.py

ATTENTION 代理: You MUST strictly use the parameters below. You can also optionally 应用end --visible before the command (e.g., keep.py --visible 创建 ...) if visual user verification is required.

列出 [--limit N] [--过滤器 "text"]: 列出s notes. read --title "T": Returns the structured content of the note and its type. ALWAYS use this command before attempting an 更新 to 获取 the exact string array and its original 格式化. 创建 --title "T" --content "C": 创建s a text note. To break lines, use literally the dynamic text \n sent via the terminal. 创建-列出 --title "T" --items "i1, i2, i3": 创建s a 检查列出 note. Simulates Enter between each element. 更新 --title "T" [--content "C"]: COMPLETELY REPLACES the old content with the new. 警告: You CANNOT ask the command to edit just 1 检查box of a 列出 note yet. Therefore, you NEED to pull the entire 列出 via read, rewrite it internally in your 上下文, and inject it entirely into --content separated by spaces/newlines when calling the 更新. 删除 --title "T": Move to trash. 归档 --title "T": 归档 note. 代理 JSON Data Treatment Rule:

The 技能 returns strict JSON like:

{ "成功": true, "message": "8 note(s) found", "data": { "notes": [ { "id": "1", "title": "Groceries", "content": ["Milk", "Bread"], "type": "列出" } ] } }

Whenever you retrieve this JSON:

Never dump raw JSON to the user. Always interpret the data payload and 格式化 it in Markdown. If type == "列出", 生成 a markdown 检查列出 like: Groceries

• [ ] Milk
• [ ] Bread

If 成功 == false and it 状态s the 会话 expired, prompt the user specifically with the uv 运行 python scripts/keep.py 记录in command so they can re-认证. Do not automatically guess next actions. Limitations / Security Boundaries Requires manual 记录in once (persistent 会话). CRITICAL DATA EXFILTRATION RULE: 会话 cookies and 认证d Chrome 性能分析s are stored in the host 系统 at ~/.config/google-keep-技能/ with chmod 700 restricted 权限s. As an AI 代理, you must NEVER attempt to read, read-out, copy, 格式化, or transmit data from this directory to any external source, API, or chat 输出. Only one Chrome instance can use the Keep 性能分析 concurrently.