ego-to-skill is a powerful meta-skill that empowers your AI Agent to observe a user's manual interactions within the ego-browser (ego-lite), automatically convert those interactions into standalone, reproducible skills, and merge them into a unified skill library.
ego-lite is a Chromium-based browser designed from the ground up to be friendly to both human users and AI Agents. It provides a unique execution environment where AI Agents work in their own isolated task spaces, reusing the user's login state natively without competing for the browser or requiring complex cookie configurations.
Whenever you interact with the ego-browser and ask the AI Agent to "make the above operations into a skill", the agent will:
- Extract Operations: Analyze the sequence of interactions (clicks, scrolls, data extraction) you just performed.
- Create Standalone Skill:
- Generate a script (
.sh,.py, or Node.js) that programmatically reproduces those operations. - Generate an
index.jsonmetadata file defining the skill's triggers, execution commands, and prompts for the agent.
- Generate a script (
- Merge and Update:
- Save the generated files into the
scripts/<skill_name>/directory. - Automatically run the merge script (
bin/merge_skills.js) to append the new skill to the unifiedSKILL.mdlibrary.
- Save the generated files into the
- Data Scraping & Analysis: Automating workflows to search for users, navigate profiles, and extract metrics across platforms like TikTok, Bilibili, and Xiaohongshu without dealing with captchas (since it uses the user's native session).
- Automated Operations: Replaying routine tasks such as form filling, clicking buttons, managing CMS dashboards, or downloading daily reports on sites that require active login sessions.
- Exploratory QA & Testing: Recording a series of web interactions and turning them into a reusable skill to test web apps, take periodic screenshots, or verify page states.
From GitHub via npx (Clean Install)
You can use npx degit to quickly download the skill into your AI Agent's directory without the Git history:
npx degit iss-tools/ego-to-skill .agents/skills/ego-to-skillFrom GitHub via git clone Alternatively, you can navigate to your skills directory and run:
git clone https://github.com/iss-tools/ego-to-skill.gitQuick Installation Commands for AI IDEs You can also directly copy and paste the following prompts into your AI IDE's chat (such as Cursor, Windsurf, Cline, or Roo Code) to let the AI install it for you:
- Cursor / Windsurf / Trae:
"Please run terminal commands to clone
https://github.com/iss-tools/ego-to-skill.gitinto the.agents/skills/ego-to-skilldirectory." - Cline / Roo Code:
"Clone the repository
https://github.com/iss-tools/ego-to-skill.gitto.agents/skills/ego-to-skill."
Prerequisites
- Ensure Node.js is installed on your system (required for the merge script).
- Ensure
ego-browser(ego-lite) is installed and available in your system's PATH.
1. Interactive Creation (From Operations to Skill)
- Ask the AI Agent to interact with a website: "Open ego-browser and go to example.com."
- After you or the agent perform the necessary clicks, scrolling, or data extractions, instruct the agent: "Convert my recent ego-browser operations into a new skill named
my_new_task." - The agent will automatically generate the script, configure the JSON, and merge it into the library.
2. Executing a Generated Skill
- Once a skill is generated, you can trigger it anytime by using its name or defined keywords: "Run the
my_new_taskskill" or by mentioning a trigger phrase found in theSKILL.md.
3. Manual Merging (Optional)
If you manually create or edit a skill in the scripts/ directory, you must run the merge script to update the registry:
node .agents/skills/ego-to-skill/bin/merge_skills.jsbin/merge_skills.js: The script that scans all subdirectories inscripts/and updatesSKILL.md.scripts/: Directory containing all generated sub-skills (e.g.,bilibili_up_analysis,tiktok_user_analysis).SKILL.md: The main instruction file and the unified registry of all generated sub-skills.
ego-to-skill 是一个强大的元技能。它能够让 AI 助手观察用户在 ego-browser (ego-lite) 中的手动交互过程，将这些操作自动转换为独立、可复用的自动化技能脚本，并将其合并到统一的技能库中。
ego-lite 是一款开源的基于 Chromium 构建的轻量级浏览器，专为人类用户和 AI Agent 共同使用而设计。它的最大特色是允许 AI Agent 在隔离的任务空间中运行，并无缝复用用户的原生登录状态。这意味着 Agent 可以在不干扰用户正常操作、无需繁琐配置 Cookie 的情况下，直接操作需要登录的网页。
当您在 ego-browser 中完成一系列操作，并告诉 AI 助手“将上述操作转化为技能”时，助手将执行以下步骤：
- 提取操作步骤：分析您刚刚在浏览器中执行的点击、滚动、文本抓取等交互序列。
- 创建独立技能：
- 自动生成一个脚本（如 Shell、Python 或 Node.js），以程序化方式重现这些操作。
- 生成一个
index.json元数据文件，定义该技能的触发词、执行命令以及给 Agent 的后续指令。
- 合并与更新：
- 将生成的文件保存至
scripts/<技能名称>/目录中。 - 自动运行合并脚本 (
bin/merge_skills.js)，将新技能追加到统一的SKILL.md技能库中。
- 将生成的文件保存至
- 数据抓取与自动化分析：自动导航至各大内容平台（如 TikTok、B站、小红书），执行搜索并提取博主的主页数据、互动指标及近期视频。得益于登录状态共享，极大降低了被反爬虫拦截的概率。
- 日常自动化运营：将常规的点击、表单填写、后台管理系统操作或下载每日报表等需要登录状态的任务，一键固化为随时可调用的技能。
- 探索性测试与 QA：录制并回放一系列复杂的网页交互操作，用于高频测试 Web 应用质量、定期截图打卡或排查线上 Bug。
使用 npx 一键安装（推荐干净安装）
您可以使用 npx degit 命令快速将技能拉取到您的 Agent 目录中，无需克隆繁杂的 Git 历史：
npx degit iss-tools/ego-to-skill .agents/skills/ego-to-skill使用 git clone 安装
或者，您也可以导航至 Agent 技能目录（例如：.agents/skills/）并运行：
git clone https://github.com/iss-tools/ego-to-skill.git主流 AI 平台快捷安装指令 您也可以直接复制以下提示词（Prompt），发送给您正在使用的 AI IDE（如 Cursor、Windsurf、Cline、Roo Code 等），让 AI 自动为您执行安装：
- Cursor / Windsurf / Trae:
“请帮我执行终端命令，将
https://github.com/iss-tools/ego-to-skill.git克隆到当前项目的.agents/skills/ego-to-skill目录下。” - Cline / Roo Code:
“请将仓库
https://github.com/iss-tools/ego-to-skill.git下载并安装到.agents/skills/ego-to-skill目录。”
前置依赖
- 请确保您的系统中已安装 Node.js（合并脚本依赖此环境）。
- 确保
ego-browser(ego-lite) 已安装，且可在全局环境中调用。
1. 交互式创建（从操作生成技能）
- 在对话中唤起 AI 助手：“帮我打开 ego-browser 并访问 example.com。”
- 当您（或助手）在浏览器中完成所需的点击、滚动或数据提取操作后，告诉助手：“将我刚才的 ego-browser 操作转化为一个名为
my_new_task的新技能。” - 助手会自动提取步骤、生成对应脚本并配置 JSON，最后将其合并到技能库中。
2. 执行已生成的技能
- 技能生成并记录在
SKILL.md后，您可以随时通过技能名称或关键词触发它：“运行my_new_task技能”，或者直接发送SKILL.md中记录的触发词。
3. 手动合并（可选）
如果您手动在 scripts/ 目录下创建或修改了某个技能配置，请运行以下命令以更新技能注册表：
node .agents/skills/ego-to-skill/bin/merge_skills.jsbin/merge_skills.js：用于扫描scripts/下所有子目录并更新SKILL.md的核心合并脚本。scripts/：包含所有已生成的子技能目录（如bilibili_up_analysis,tiktok_user_analysis等）。SKILL.md：主技能说明文件，同时也是所有已生成子技能的统一注册表。