agentscope-ai/skills

# Documentation and Source Code Quick Reference

When answering questions about **installation, configuration, or behavioral principles**, first **classify by keyword**, then **open 1–2 paths most likely to contain the answer** from the table below, avoiding aimless directory traversal.

## Usage Steps

1. Extract the topic from the user's question (match against the left column or synonyms in the table below).
2. Resolve **`$QWENPAW_ROOT`**: use `which qwenpaw` to get the executable path. If it is `…/.qwenpaw/bin/qwenpaw`, the source root is three levels up (consistent with the **guidance** skill); otherwise, determine it from the user-provided installation path.
3. **Read documentation first**: `website/public/docs/<topic>.<language>.md` (use the same language as the user: `zh` / `en` / `ru`, etc.). If that is insufficient, read the **source entry points** listed in the table.

## Topic / Keywords → Preferred Documentation and Source Code

| Topic or Keywords (examples) | Preferred Documentation (`website/public/docs/`) | Common Source Entry Points (relative to `$QWENPAW_ROOT`) |
|---------------------|-----------------------------------|-----------------------------------|

# 文档与源码速查

回答 **安装、配置、行为原理** 类问题时，先 **按关键词归类**，再按下表 **打开 1～2 个最可能命中的路径** 阅读，避免长时间无目的遍历。

## 使用步骤

1. 从用户问题中提取主题（对照下表左列或同类词）。
2. 解析 **`$QWENPAW_ROOT`**：以 `which qwenpaw` 得到可执行路径，若为 `…/.qwenpaw/bin/qwenpaw` 则源码根为其上三级目录（与 **guidance** skill 一致）；否则结合用户给出的安装路径判断。
3. **先读文档** `website/public/docs/<专题>.<语言>.md`（语言取与用户一致：`zh` / `en` / `ru` 等），仍不足再读表中 **源码入口**。

## 主题 / 关键词 → 优先文档与源码

| 主题或关键词（示例） | 优先文档（`website/public/docs/`） | 常见源码入口（相对 `$QWENPAW_ROOT`） |
|---------------------|-----------------------------------|-----------------------------------|

# AgentScope 2.0

AgentScope is an open-source framework for building and serving LLM-powered
agent applications, from a single tool-using agent to coordinated multi-agent
systems. It provides application orchestration and service infrastructure;
model inference comes from configured providers, and tools execute through
configured local or sandbox backends. It consists of two layers:

- **Agent SDK:** Building blocks for agent applications, including agents,
  models, messages, tools, context and state management, middleware, memory,
  RAG, multi-agent orchestration, and workspaces.
- **Service:** A service layer built on the SDK, providing APIs for agent and
  session management, persistence, teams, scheduling, channels, and resource
  management, with a Web UI example.

# Browser CDP Reference

By default, **browser_use** launches and manages the local Chrome/Chromium via **managed CDP**, but this does not mean the CDP port should be exposed to the user or other tools every time.

This skill focuses on more "explicit" CDP usage:

1. **Scanning local CDP ports**
2. **Connecting to an existing Chrome (`connect_cdp`)**
3. **Explicitly specifying a `cdp_port` when launching the browser**
4. **Sharing a single browser instance across multiple agents/tools**

In other words:

- The default `start` uses managed CDP under the hood, but users typically do not need to understand or be aware of CDP details

# 浏览器 CDP 使用参考

当前 **browser_use** 默认就是以 **managed CDP** 方式启动并接管本地 Chrome/Chromium，但这不等于每次都应该把 CDP 端口暴露给用户或其他工具。

本 skill 关注的是更“显式”的 CDP 用法：

1. **扫描本地 CDP 端口**
2. **连接已有 Chrome（`connect_cdp`）**
3. **显式指定 `cdp_port` 启动浏览器**
4. **让多个 agent / 工具共享同一个浏览器实例**

也就是说：

- 普通 `start` 默认会使用 managed CDP，但通常不需要用户理解或感知 CDP 细节

# Browser Launch Modes

`browser_use.start` has only two launch modes:

- Default: managed CDP
- `private_mode=true`: Playwright-managed

Parameter meanings:

- `headed`: whether to display the browser window
- `private_mode`: whether to disable CDP and use Playwright instead

The two parameters are independent and can be freely combined.

# 浏览器启动模式

`browser_use.start` 只有两种启动方式：

- 默认：managed CDP
- `private_mode=true`：Playwright 直接管理

参数含义：

- `headed`：是否显示浏览器窗口
- `private_mode`：是否禁用 CDP，改走 Playwright
- `browser_args`：额外的 Chromium 启动参数（字符串），多个参数用空格分隔。适用于所有启动路径（headless、headed、managed CDP）。例如 `"--incognito"` 启用隐身模式，`"--proxy-server=http://127.0.0.1:7890"` 设置代理。默认空字符串（无额外参数）。
- `executable_path`：自定义浏览器可执行文件路径（字符串）。设置后覆盖系统默认浏览器检测，可指定任意基于 Chromium 的浏览器。例如 `"C:\\Program Files\\Google\\Chrome\\Application\\chrome.exe"`。仅允许包含已知浏览器关键词（chrome、chromium、edge、firefox、brave 等）的可执行文件，且文件必须存在。默认空字符串（使用系统默认）。

# Channel Message

## When to Use

Use this skill only when the **user explicitly asks you to send a message to a channel/session**, or when you need to **proactively push a notification** (e.g., task completion, reminders, alerts).
This is a **one-way send** — **no reply is returned**.

### Should Use
- The user explicitly asks to send to a specific channel/session
- Proactively notifying the user after a task completes
- Scheduled reminders, alerts, or status updates
- Pushing async results back to an existing session
- The user explicitly says "notify me when done"

# 频道消息推送

## 什么时候用

通常只有在**用户明确要求向某个 channel / 会话发送消息**，或你需要**主动推送通知**（如任务完成、提醒、告警）时，使用本 skill。  
这是**单向发送**，**不会返回回复**。

### 应该使用
- 用户明确要求发往某个 channel / 会话
- 任务完成后主动通知用户
- 定时提醒、告警、状态更新
- 将异步结果推送回某个已有会话
- 用户明确要求"处理完后通知我"

# Chat with Agent

## When to Use

Use this skill when you need to **ask another agent a question, seek help, request a plan, request a review, request decision support, or engage in any form of communication**.
If the **user explicitly asks to talk to a specific agent**, you should also use this skill.

### Should Use

- You need another agent's expertise, judgment, or second opinion
- You need to request a plan, review, or recommendation from an agent
- The user explicitly asks a specific agent to participate, assist, or answer
- You need to continue an existing agent session while preserving context

# 与 Agent 对话

## 什么时候用

当你需要**向另一个 agent 询问问题、寻求帮助、请求方案、请求复核、请求决策支持，或者进行任何形式的交流**时，使用本 skill。  
如果**用户明确要求与某个 agent 对话**，也应使用本 skill。

### 应该使用

- 需要另一个 agent 的专长、判断或第二意见
- 需要向某个 agent 请求方案、复核或建议
- 用户明确要求某个 agent 参与、协助或回答
- 需要继续某个已有的 agent 会话，并保留上下文

# Cron (Scheduled Task Management)

## When to Use

Use this skill only when you need to **automatically execute something at a future time** or **repeat execution on a schedule**.

### Should Use
- User asks to do something "daily / weekly / hourly"
- User asks for automatic reminders or execution "tomorrow at 9 AM / next Monday / at a specific time"
- Long-term periodic notifications, checks, or reports are needed

### Should Not Use
- The task only needs to be **executed once right now**
- It is just a normal reply within the current session

# 定时任务管理

## 什么时候用

只有在需要**未来某个时间自动执行**，或**按周期重复执行**时，使用本 skill。

### 应该使用
- 用户要求"每天 / 每周 / 每小时"执行某事
- 用户要求"明天 9 点 / 下周一 / 某个时间"自动提醒或执行
- 需要长期周期性通知、检查、汇报

### 不应使用
- 只是要**现在立即执行一次**
- 只是当前会话中的正常回复

# DingTalk Channel Auto-Connect (Headed Browser)

This skill automates the creation of a DingTalk application and the binding of a QwenPaw channel using a headed browser.

## Mandatory Rules

1. Must launch in headed browser mode:

```json
{"action": "start", "headed": true}
```

2. Must pause when a login gate is encountered:
   - If the page displays a login screen (e.g., login prompt, QR code login, phone/password login), stop automated operations immediately.

# 钉钉频道自动连接（可视浏览器）

此 skill 用于通过可视浏览器自动化完成钉钉应用创建与 QwenPaw channel 绑定。

## 强制规则

1. 必须使用可视浏览器模式启动：

```json
{"action": "start", "headed": true}
```

2. 遇到登录关卡必须暂停：
   - 若页面出现登录界面（如 `登录`、扫码登录、手机号/密码登录），立即停止自动操作。

# File Reader Toolbox

Use this skill when the user asks to read or summarize local text-based files. PDFs, Office documents, images, audio, and video are out of scope for this skill and should be handled by their dedicated skills/tools.

## Quick Type Check

Use a type probe before reading:

```bash
file -b --mime-type "/path/to/file"
```

If the file is large, avoid dumping the whole content; extract a small, relevant portion and summarize.

## Text-Based Files (use read_file)

# 文件读取工具箱

当用户要求读取或总结本地文本文件时使用此技能。PDF、Office 文档、图片、音频和视频不在此技能范围内，应由其专用技能/工具处理。

## 快速类型检查

读取前先进行类型探测:

```bash
file -b --mime-type "/path/to/file"
```

如果文件较大，避免输出全部内容；提取小部分相关内容并进行总结。

## 文本文件（使用 read_file）

# QwenPaw Installation and Configuration Q&A Guide

Use this skill when the user asks about **QwenPaw installation, initialization, environment configuration, dependency requirements, or common configuration options**.

Core principles:

- Check local documentation first, then answer
- Base answers on what has actually been read, do not speculate
- Answer in the same language the user used to ask

## Standard Flow

### Step 1: Locate the Documentation Directory

# QwenPaw 安装与配置问答指南

当用户询问 **QwenPaw 的安装、初始化、环境配置、依赖要求、常见配置项** 时，使用本 skill。

核心原则：

- 先查本地文档，再回答
- 回答要基于已读到的内容，不臆测
- 回答语言与用户提问语言保持一致

## 标准流程


### 第一步：定位文档位置

# Himalaya Email CLI

Himalaya is a CLI email client that lets you manage emails from the terminal using IMAP, SMTP, Notmuch, or Sendmail backends.

## References

- `references/configuration.md` (config file setup + IMAP/SMTP authentication)

## Prerequisites

1. **Himalaya CLI** - the `himalaya` binary must already be on `PATH`. Check with `himalaya --version`.
   - **Recommended: v1.2.0 or newer.** Older releases can fail against some IMAP servers; v1.2.0+ includes related fixes.
2. A configuration file at `~/.config/himalaya/config.toml`
3. IMAP/SMTP credentials configured (password stored securely)

# Himalaya 邮件命令行工具

Himalaya 是一个命令行邮件客户端，可以通过 IMAP、SMTP、Notmuch 或 Sendmail 后端在终端中管理邮件。

## 参考资料

- `references/configuration.md`（配置文件设置 + IMAP/SMTP 认证）

## 前置条件

1. **Himalaya CLI** - `himalaya` 二进制文件必须已在 `PATH` 中。通过 `himalaya --version` 检查。
   - **推荐使用 v1.2.0 或更新版本。** 旧版本在某些 IMAP 服务器上可能会失败；v1.2.0+ 包含了相关修复。
2. 配置文件位于 `~/.config/himalaya/config.toml`
3. 已配置 IMAP/SMTP 凭据（密码安全存储）

# Make Plan

Use this Skill when you need to make an **external plan request** to a stronger Agent.

The goal of this Skill is not to outsource a task, but rather to:
- Request a plan from a stronger Agent
- The plan must consist of **clear, actionable steps**
- **You execute the plan yourself**
- Do not ask the consulted Agent to execute the task directly

How to invoke:
- Use `list_agents()` to check available Agents
- Use `chat_with_agent(...)` to request the stronger Agent to "make a plan"
- In `text`, explicitly include the prompt: **only provide a step-by-step executable plan, do not execute**

# 制定计划

当你需要向更强 Agent 发起一次**外部计划请求**时，使用本 Skill。

本 Skill 的目标不是把任务外包出去，而是：
- 向更强 Agent 要一份计划
- 这份计划必须是**明确的执行步骤**
- **由你自己按计划执行**
- 不要求被求助的 Agent 直接动手执行任务

调用步骤：
1) 使用 `list_agents()` 查可用 Agent 的 ID
  - 可根据返回的描述信息选择最合适的 Agent 来求助
  - 如果找不到非常合适的 Agent，就选择 default

# Multi-Agent Collaboration

## When to Use

Use this skill when you **need another agent's expertise, context, workspace content, or collaborative support**.
If the **user explicitly asks a specific agent to participate/assist/answer**, you should also use this skill.

### Should Use
- The current task is clearly better suited for a specialized agent
- You need another agent's workspace / files / context
- You need a second opinion or expert review
- The user explicitly asks for a specific agent to participate or to invoke another agent

### Should Not Use

# 多智能体协作

## 什么时候用

当你**需要其他 agent 的专业能力、上下文、workspace 内容或协作支持**时，使用本 skill。  
如果**用户明确要求某个 agent 参与/协助/回答**，也应使用本 skill。

### 应该使用
- 当前任务明显更适合某个专用 agent
- 需要另一个 agent 的 workspace / 文件 / 上下文
- 需要第二意见或专业复核
- 用户明确要求某个 agent 参与或调用其他 agent

### 不应使用

## 🧠 Memory Architecture

Each session is stateless. Your memory lives entirely within the local file system. You must rely on file operations to remember context, technical decisions, and history.

### Core Memory Files
- **`memory/YYYY-MM-DD.md` (Daily Logs):** Raw, chronological logs of what happened. Use this for daily tasks, scratchpad thinking, and immediate context.
- **`MEMORY.md` (Long-Term Memory):** Your curated, distilled knowledge base. Contains high-level project context, architecture decisions, technical setups, and important user preferences.

## 🛠<U+FE0F> Core Memory Operations

You have access to native file tools and standard OS commands. Use them strictly in these patterns to avoid data loss or hallucinations:

### 1. 🔍 Search (OS-Specific Commands)
- **When to use:** ALWAYS use this before answering questions about past work, decisions, or dates to find where information is stored.

# News Reference

When the user asks for "latest news", "what's in the news today", or "news in category X", use the **browser_use** tool with the categories and URLs below: open the page, take a snapshot, then extract headlines and key points from the page content and reply to the user.

## Categories and Sources

| Category      | Source                    | URL |
|---------------|---------------------------|-----|
| **Politics**  | People's Daily · CPC News | https://cpc.people.com.cn/ |
| **Finance**   | China Economic Net        | http://www.ce.cn/ |
| **Society**   | China News · Society      | https://www.chinanews.com/society/ |
| **World**     | CGTN                      | https://www.cgtn.com/ |
| **Tech**      | Science and Technology Daily | https://www.stdaily.com/ |
| **Sports**    | CCTV Sports               | https://sports.cctv.com/ |

# 新闻参考

当用户询问"最新新闻"、"今天有什么新闻"或"某某类别的新闻"时，使用 **browser_use** 工具配合以下类别和 URL：打开页面，截取快照，然后从页面内容中提取标题和要点并回复用户。

## 类别和来源

| 类别      | 来源                    | URL |
|-----------|--------------------------|-----|
| **政治**  | 人民网 · 中国共产党新闻网 | https://cpc.people.com.cn/ |
| **财经**  | 中国经济网               | http://www.ce.cn/ |
| **社会**  | 中新网 · 社会            | https://www.chinanews.com/society/ |
| **国际**  | CGTN                     | https://www.cgtn.com/ |
| **科技**  | 科技日报                 | https://www.stdaily.com/ |
| **体育**  | 央视体育                 | https://sports.cctv.com/ |