# qwen-agent-docs - Doramagic AI Context Pack

> 定位：安装前体验与判断资产。它帮助宿主 AI 有一个好的开始，但不代表已经安装、执行或验证目标项目。

## 充分原则

- **充分原则，不是压缩原则**：AI Context Pack 应该充分到让宿主 AI 在开工前理解项目价值、能力边界、使用入口、风险和证据来源；它可以分层组织，但不以最短摘要为目标。
- **压缩策略**：只压缩噪声和重复内容，不压缩会影响判断和开工质量的上下文。

## 给宿主 AI 的使用方式

你正在读取 Doramagic 为 qwen-agent-docs 编译的 AI Context Pack。请把它当作开工前上下文：帮助用户理解适合谁、能做什么、如何开始、哪些必须安装后验证、风险在哪里。不要声称你已经安装、运行或执行了目标项目。

## Claim 消费规则

- **事实来源**：Repo Evidence + Claim/Evidence Graph；Human Wiki 只提供显著性、术语和叙事结构。
- **事实最低状态**：`supported`
- `supported`：可以作为项目事实使用，但回答中必须引用 claim_id 和证据路径。
- `weak`：只能作为低置信度线索，必须要求用户继续核实。
- `inferred`：只能用于风险提示或待确认问题，不能包装成项目事实。
- `unverified`：不得作为事实使用，应明确说证据不足。
- `contradicted`：必须展示冲突来源，不得替用户强行选择一个版本。

## 它最适合谁

- **想在安装前理解开源项目价值和边界的用户**：当前证据主要来自项目文档。 证据：`README.md` Claim：`clm_0002` supported 0.86

## 它能做什么

- **命令行启动或安装流程**（需要安装后验证）：项目文档中存在可执行命令，真实使用需要在本地或宿主环境中运行这些命令。 证据：`README.md` Claim：`clm_0001` supported 0.86

## 怎么开始

- `pip install -U "qwen-agent[gui,rag,code_interpreter,mcp]"` 证据：`README.md` Claim：`clm_0003` supported 0.86
- `git clone https://github.com/QwenLM/Qwen-Agent.git` 证据：`README.md` Claim：`clm_0004` supported 0.86
- `pip install -e ./"[gui,rag,code_interpreter,mcp]"` 证据：`README.md` Claim：`clm_0005` supported 0.86

## 继续前判断卡

- **当前建议**：需要管理员/安全审批
- **为什么**：继续前可能涉及密钥、账号、外部服务或敏感上下文，建议先经过管理员或安全审批。

### 30 秒判断

- **现在怎么做**：需要管理员/安全审批
- **最小安全下一步**：先跑 Prompt Preview；若涉及凭证或企业环境，先审批再试装
- **先别相信**：角色质量和任务匹配不能直接相信。
- **继续会触碰**：角色选择偏差、命令执行、本地环境或项目文件

### 现在可以相信

- **适合人群线索：想在安装前理解开源项目价值和边界的用户**（supported）：有 supported claim 或项目证据支撑，但仍不等于真实安装效果。 证据：`README.md` Claim：`clm_0002` supported 0.86
- **能力存在：命令行启动或安装流程**（supported）：可以相信项目包含这类能力线索；是否适合你的具体任务仍要试用或安装后验证。 证据：`README.md` Claim：`clm_0001` supported 0.86
- **存在 Quick Start / 安装命令线索**（supported）：可以相信项目文档出现过启动或安装入口；不要因此直接在主力环境运行。 证据：`README.md` Claim：`clm_0003` supported 0.86

### 现在还不能相信

- **角色质量和任务匹配不能直接相信。**（unverified）：角色库证明有很多角色，不证明每个角色都适合你的具体任务，也不证明角色能产生高质量结果。
- **不能把角色文案当成真实执行能力。**（unverified）：安装前只能判断角色描述和任务画像是否匹配，不能证明它能在宿主 AI 里完成任务。
- **真实输出质量不能在安装前相信。**（unverified）：Prompt Preview 只能展示引导方式，不能证明真实项目中的结果质量。
- **宿主 AI 版本兼容性不能在安装前相信。**（unverified）：Claude、Cursor、Codex、Gemini 等宿主加载规则和版本差异必须在真实环境验证。
- **不会污染现有宿主 AI 行为，不能直接相信。**（inferred）：Skill、plugin、AGENTS/CLAUDE/GEMINI 指令可能改变宿主 AI 的默认行为。
- **可安全回滚不能默认相信。**（unverified）：除非项目明确提供卸载和恢复说明，否则必须先在隔离环境验证。
- **真实安装后是否与用户当前宿主 AI 版本兼容？**（unverified）：兼容性只能通过实际宿主环境验证。
- **项目输出质量是否满足用户具体任务？**（unverified）：安装前预览只能展示流程和边界，不能替代真实评测。

### 继续会触碰什么

- **角色选择偏差**：用户对任务应该由哪个专家角色处理的判断。 原因：选错角色会让 AI 从错误专业视角回答，浪费时间或误导决策。
- **命令执行**：包管理器、网络下载、本地插件目录、项目配置或用户主目录。 原因：运行第一条命令就可能产生环境改动；必须先判断是否值得跑。 证据：`README.md`
- **本地环境或项目文件**：安装结果、插件缓存、项目配置或本地依赖目录。 原因：安装前无法证明写入范围和回滚方式，需要隔离验证。 证据：`README.md`
- **环境变量 / API Key**：项目入口文档明确出现 API key、token、secret 或账号凭证配置。 原因：如果真实安装需要凭证，应先使用测试凭证并经过权限/合规判断。 证据：`README.md`, `README_CN.md`, `benchmark/deepplanning/README.md`, `benchmark/deepplanning/models_config.json` 等
- **宿主 AI 上下文**：AI Context Pack、Prompt Preview、Skill 路由、风险规则和项目事实。 原因：导入上下文会影响宿主 AI 后续判断，必须避免把未验证项包装成事实。

### 最小安全下一步

- **先跑 Prompt Preview**：先用交互式试用验证任务画像和角色匹配，不要先导入整套角色库。（适用：任何项目都适用，尤其是输出质量未知时。）
- **只在隔离目录或测试账号试装**：避免安装命令污染主力宿主 AI、真实项目或用户主目录。（适用：存在命令执行、插件配置或本地写入线索时。）
- **不要使用真实生产凭证**：环境变量/API key 一旦进入宿主或工具链，可能产生账号和合规风险。（适用：出现 API、TOKEN、KEY、SECRET 等环境线索时。）
- **安装后只验证一个最小任务**：先验证加载、兼容、输出质量和回滚，再决定是否深用。（适用：准备从试用进入真实工作流时。）

### 退出方式

- **保留安装前状态**：记录原始宿主配置和项目状态，后续才能判断是否可恢复。
- **保留原始角色选择记录**：如果输出偏题，可以回到任务画像阶段重新选择角色，而不是继续沿着错误角色推进。
- **记录安装命令和写入路径**：没有明确卸载说明时，至少要知道哪些目录或配置需要手动清理。
- **准备撤销测试 API key 或 token**：测试凭证泄露或误用时，可以快速止损。
- **如果没有回滚路径，不进入主力环境**：不可回滚是继续前阻断项，不应靠信任或运气继续。

## 哪些只能预览

- 解释项目适合谁和能做什么
- 基于项目文档演示典型对话流程
- 帮助用户判断是否值得安装或继续研究

## 哪些必须安装后验证

- 真实安装 Skill、插件或 CLI
- 执行脚本、修改本地文件或访问外部服务
- 验证真实输出质量、性能和兼容性

## 边界与风险判断卡

- **把安装前预览误认为真实运行**：用户可能高估项目已经完成的配置、权限和兼容性验证。 处理方式：明确区分 prompt_preview_can_do 与 runtime_required。 Claim：`clm_0006` inferred 0.45
- **命令执行会修改本地环境**：安装命令可能写入用户主目录、宿主插件目录或项目配置。 处理方式：先在隔离环境或测试账号中运行。 证据：`README.md` Claim：`clm_0007` supported 0.86
- **待确认**：真实安装后是否与用户当前宿主 AI 版本兼容？。原因：兼容性只能通过实际宿主环境验证。
- **待确认**：项目输出质量是否满足用户具体任务？。原因：安装前预览只能展示流程和边界，不能替代真实评测。
- **待确认**：安装命令是否需要网络、权限或全局写入？。原因：这影响企业环境和个人环境的安装风险。

## 开工前工作上下文

### 加载顺序

- 先读取 how_to_use.host_ai_instruction，建立安装前判断资产的边界。
- 读取 claim_graph_summary，确认事实来自 Claim/Evidence Graph，而不是 Human Wiki 叙事。
- 再读取 intended_users、capabilities 和 quick_start_candidates，判断用户是否匹配。
- 需要执行具体任务时，优先查 role_skill_index，再查 evidence_index。
- 遇到真实安装、文件修改、网络访问、性能或兼容性问题时，转入 risk_card 和 boundaries.runtime_required。

### 任务路由

- **命令行启动或安装流程**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`README.md` Claim：`clm_0001` supported 0.86

### 上下文规模

- 文件总数：281
- 重要文件覆盖：40/281
- 证据索引条目：77
- 角色 / Skill 条目：22

### 证据不足时的处理

- **missing_evidence**：说明证据不足，要求用户提供目标文件、README 段落或安装后验证记录；不要补全事实。
- **out_of_scope_request**：说明该任务超出当前 AI Context Pack 证据范围，并建议用户先查看 Human Manual 或真实安装后验证。
- **runtime_request**：给出安装前检查清单和命令来源，但不要替用户执行命令或声称已执行。
- **source_conflict**：同时展示冲突来源，标记为待核实，不要强行选择一个版本。

## Prompt Recipes

### 适配判断

- 目标：判断这个项目是否适合用户当前任务。
- 预期输出：适配结论、关键理由、证据引用、安装前可预览内容、必须安装后验证内容、下一步建议。

```text
请基于 qwen-agent-docs 的 AI Context Pack，先问我 3 个必要问题，然后判断它是否适合我的任务。回答必须包含：适合谁、能做什么、不能做什么、是否值得安装、证据来自哪里。所有项目事实必须引用 evidence_refs、source_paths 或 claim_id。
```

### 安装前体验

- 目标：让用户在安装前感受核心工作流，同时避免把预览包装成真实能力或营销承诺。
- 预期输出：一段带边界标签的体验剧本、安装后验证清单和谨慎建议；不含真实运行承诺或强营销表述。

```text
请把 qwen-agent-docs 当作安装前体验资产，而不是已安装工具或真实运行环境。

请严格输出四段：
1. 先问我 3 个必要问题。
2. 给出一段“体验剧本”：用 [安装前可预览]、[必须安装后验证]、[证据不足] 三种标签展示它可能如何引导工作流。
3. 给出安装后验证清单：列出哪些能力只有真实安装、真实宿主加载、真实项目运行后才能确认。
4. 给出谨慎建议：只能说“值得继续研究/试装”“先补充信息后再判断”或“不建议继续”，不得替项目背书。

硬性边界：
- 不要声称已经安装、运行、执行测试、修改文件或产生真实结果。
- 不要写“自动适配”“确保通过”“完美适配”“强烈建议安装”等承诺性表达。
- 如果描述安装后的工作方式，必须使用“如果安装成功且宿主正确加载 Skill，它可能会……”这种条件句。
- 体验剧本只能写成“示例台词/假设流程”：使用“可能会询问/可能会建议/可能会展示”，不要写“已写入、已生成、已通过、正在运行、正在生成”。
- Prompt Preview 不负责给安装命令；如用户准备试装，只能提示先阅读 Quick Start 和 Risk Card，并在隔离环境验证。
- 所有项目事实必须来自 supported claim、evidence_refs 或 source_paths；inferred/unverified 只能作风险或待确认项。

```

### 角色 / Skill 选择

- 目标：从项目里的角色或 Skill 中挑选最匹配的资产。
- 预期输出：候选角色或 Skill 列表，每项包含适用场景、证据路径、风险边界和是否需要安装后验证。

```text
请读取 role_skill_index，根据我的目标任务推荐 3-5 个最相关的角色或 Skill。每个推荐都要说明适用场景、可能输出、风险边界和 evidence_refs。
```

### 风险预检

- 目标：安装或引入前识别环境、权限、规则冲突和质量风险。
- 预期输出：环境、权限、依赖、许可、宿主冲突、质量风险和未知项的检查清单。

```text
请基于 risk_card、boundaries 和 quick_start_candidates，给我一份安装前风险预检清单。不要替我执行命令，只说明我应该检查什么、为什么检查、失败会有什么影响。
```

### 宿主 AI 开工指令

- 目标：把项目上下文转成一次对话开始前的宿主 AI 指令。
- 预期输出：一段边界明确、证据引用明确、适合复制给宿主 AI 的开工前指令。

```text
请基于 qwen-agent-docs 的 AI Context Pack，生成一段我可以粘贴给宿主 AI 的开工前指令。这段指令必须遵守 not_runtime=true，不能声称项目已经安装、运行或产生真实结果。
```


## 角色 / Skill 索引

- 共索引 22 个角色 / Skill / 项目文档条目。

- **News**（project_doc）：<!--- Copyright 2023 The Qwen team, Alibaba Group. All rights reserved. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`README.md`
- **Code Interpreter Benchmark**（project_doc）：Introduction To assess LLM's ability to use the Python Code Interpreter for tasks such as mathematical problem solving, data visualization, and other general-purpose tasks such as file handling and web scraping, we have created and open-sourced a benchmark specifically designed for evaluating these capabilities. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`benchmark/code_interpreter/README.md`
- **DeepPlanning Benchmark**（project_doc）：A comprehensive benchmark for evaluating AI agents' planning capabilities across multiple domains. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`benchmark/deepplanning/README.md`
- **🛠️ Quick Start**（project_doc）：This domain can be run as part of the unified benchmark or independently. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`benchmark/deepplanning/shoppingplanning/README.md`
- **🛠️ Quick Start**（project_doc）：This domain can be run as part of the unified benchmark or independently. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`benchmark/deepplanning/travelplanning/README.md`
- **Installation**（project_doc）：- Install the stable version from PyPI: 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`qwen-agent-docs/website/content/en/guide/get_started/install.md`
- **更新**（project_doc）：<!--- Copyright 2023 The Qwen team, Alibaba Group. All rights reserved. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`README_CN.md`
- **Example Application: BrowserQwen**（project_doc）：<!--- Copyright 2023 The Qwen team, Alibaba Group. All rights reserved. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`browser_qwen.md`
- **示例应用：BrowserQwen**（project_doc）：<!--- Copyright 2023 The Qwen team, Alibaba Group. All rights reserved. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`browser_qwen_cn.md`
- **Benchmark Overview**（project_doc）：We provide a benchmark to evaluate the planning capabilities of state-of-the-art agentic models. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`qwen-agent-docs/website/content/en/benchmarks/index.md`
- **Agent Introduction**（project_doc）：This document introduces the usage and development process of the Agent class. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`qwen-agent-docs/website/content/en/guide/core_moduls/agent.md`
- **Context Management**（project_doc）：The context management logic of Qwen Agent aims to dynamically truncate input messages while maintaining the rationality of the dialogue structure, so that the total number of tokens does not exceed the maximum context length supported by the model. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`qwen-agent-docs/website/content/en/guide/core_moduls/context.md`
- **LLM Introduction**（project_doc）：This document introduces the usage and development process of LLM classes. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`qwen-agent-docs/website/content/en/guide/core_moduls/llm.md`
- **MCP Model Context Protocol**（project_doc）：MCP Model Context Protocol is a standardized protocol that enables large language models LLMs to interact with external tools and services in a structured way. In the Qwen-Agent framework, MCP is deeply integrated to empower intelligent agents with capabilities such as file system access, memory management, database queries, and more. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`qwen-agent-docs/website/content/en/guide/core_moduls/mcp.md`
- **RAG Retrieval-Augmented Generation**（project_doc）：Qwen-Agent provides built-in RAG Retrieval-Augmented Generation capabilities to enhance responses by retrieving relevant content from a given set of documents. This allows the language model to ground its answers in specific, user-provided knowledge sources. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`qwen-agent-docs/website/content/en/guide/core_moduls/rag.md`
- **Qwen-Agent Schema Documentation**（project_doc）：The qwen-agent schema provides a structured, type-safe messaging system to support advanced capabilities such as multimodal conversations, function calling, and reasoning chains. Built on Pydantic, it ensures data integrity during construction, validation, and serialization while enabling flexible representation of multimodal content text, images, files, audio, and video . 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`qwen-agent-docs/website/content/en/guide/core_moduls/schema.md`
- **Tool Introduction**（project_doc）：This document introduces the usage and development process of the Tool class. Please refer to the Configuration ../get started/configuration.md for detailed parameter configuration. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`qwen-agent-docs/website/content/en/guide/core_moduls/tool.md`
- **Configuration**（project_doc）：This document explains all configuration parameters of Agent. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`qwen-agent-docs/website/content/en/guide/get_started/configuration.md`
- **Qwen-Agent Features**（project_doc）：Qwen-Agent is a powerful and flexible framework for building intelligent LLM-powered applications. Key features include: 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`qwen-agent-docs/website/content/en/guide/get_started/features.md`
- **QuickStart**（project_doc）：This quickstart will guide you to implement an agent using a few lines of code in just a few minutes. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`qwen-agent-docs/website/content/en/guide/get_started/quickstart.md`
- **Qwen-Agent Overview**（project_doc）：Qwen-Agent is a framework for developing LLM applications based on the instruction following, tool usage, planning, and memory capabilities of Qwen. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`qwen-agent-docs/website/content/en/guide/index.md`
- **Qwen Agent Documentation**（project_doc）：Welcome to the Qwen Agent documentation. Qwen Agent is an open-source AI agent framework that helps you build powerful LLM applications based on the instruction following, tool usage, planning, and memory capabilities of Qwen. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`qwen-agent-docs/website/content/en/index.md`

## 证据索引

- 共索引 77 条证据。

- **News**（documentation）：<!--- Copyright 2023 The Qwen team, Alibaba Group. All rights reserved. 证据：`README.md`
- **Code Interpreter Benchmark**（documentation）：Introduction To assess LLM's ability to use the Python Code Interpreter for tasks such as mathematical problem solving, data visualization, and other general-purpose tasks such as file handling and web scraping, we have created and open-sourced a benchmark specifically designed for evaluating these capabilities. 证据：`benchmark/code_interpreter/README.md`
- **DeepPlanning Benchmark**（documentation）：A comprehensive benchmark for evaluating AI agents' planning capabilities across multiple domains. 证据：`benchmark/deepplanning/README.md`
- **🛠️ Quick Start**（documentation）：This domain can be run as part of the unified benchmark or independently. 证据：`benchmark/deepplanning/shoppingplanning/README.md`
- **🛠️ Quick Start**（documentation）：This domain can be run as part of the unified benchmark or independently. 证据：`benchmark/deepplanning/travelplanning/README.md`
- **Package**（package_manifest）：{ "name": "qwen-agent-docs", "version": "1.0.0", "description": "Documentation for Qwen-Agent", "scripts": { "clean": "rm -rf .next", "dev": "npm run clean && next dev", "build": "npm run clean && next build" }, "keywords": "qwen", "code", "documentation", "translation", "ai", "nextra" , "author": "Qwen Agent Team", "license": "MIT", "dependencies": { "@radix-ui/react-alert-dialog": "^1.1.14", "@radix-ui/react-avatar": "^1.1.10", "@radix-ui/react-checkbox": "^1.1.11", "@radix-ui/react-dialog": "^1.1.14", "@radix-ui/react-dropdown-menu": "^2.1.15", "@radix-ui/react-label": "^2.1.7", "@radix-ui/react-navigation-menu": "^1.2.13", "@radix-ui/react-popover": "^1.1.14", "@radix-ui/react-progress"… 证据：`qwen-agent-docs/website/package.json`
- **Installation**（documentation）：- Install the stable version from PyPI: 证据：`qwen-agent-docs/website/content/en/guide/get_started/install.md`
- **License**（source_file）：Apache License Version 2.0, January 2004 http://www.apache.org/licenses/ 证据：`LICENSE`
- **更新**（documentation）：<!--- Copyright 2023 The Qwen team, Alibaba Group. All rights reserved. 证据：`README_CN.md`
- **Example Application: BrowserQwen**（documentation）：<!--- Copyright 2023 The Qwen team, Alibaba Group. All rights reserved. 证据：`browser_qwen.md`
- **示例应用：BrowserQwen**（documentation）：<!--- Copyright 2023 The Qwen team, Alibaba Group. All rights reserved. 证据：`browser_qwen_cn.md`
- **Benchmark Overview**（documentation）：We provide a benchmark to evaluate the planning capabilities of state-of-the-art agentic models. 证据：`qwen-agent-docs/website/content/en/benchmarks/index.md`
- **Agent Introduction**（documentation）：This document introduces the usage and development process of the Agent class. 证据：`qwen-agent-docs/website/content/en/guide/core_moduls/agent.md`
- **Context Management**（documentation）：The context management logic of Qwen Agent aims to dynamically truncate input messages while maintaining the rationality of the dialogue structure, so that the total number of tokens does not exceed the maximum context length supported by the model. 证据：`qwen-agent-docs/website/content/en/guide/core_moduls/context.md`
- **LLM Introduction**（documentation）：This document introduces the usage and development process of LLM classes. 证据：`qwen-agent-docs/website/content/en/guide/core_moduls/llm.md`
- **MCP Model Context Protocol**（documentation）：MCP Model Context Protocol is a standardized protocol that enables large language models LLMs to interact with external tools and services in a structured way. In the Qwen-Agent framework, MCP is deeply integrated to empower intelligent agents with capabilities such as file system access, memory management, database queries, and more. 证据：`qwen-agent-docs/website/content/en/guide/core_moduls/mcp.md`
- **RAG Retrieval-Augmented Generation**（documentation）：Qwen-Agent provides built-in RAG Retrieval-Augmented Generation capabilities to enhance responses by retrieving relevant content from a given set of documents. This allows the language model to ground its answers in specific, user-provided knowledge sources. 证据：`qwen-agent-docs/website/content/en/guide/core_moduls/rag.md`
- **Qwen-Agent Schema Documentation**（documentation）：The qwen-agent schema provides a structured, type-safe messaging system to support advanced capabilities such as multimodal conversations, function calling, and reasoning chains. Built on Pydantic, it ensures data integrity during construction, validation, and serialization while enabling flexible representation of multimodal content text, images, files, audio, and video . 证据：`qwen-agent-docs/website/content/en/guide/core_moduls/schema.md`
- **Tool Introduction**（documentation）：This document introduces the usage and development process of the Tool class. Please refer to the Configuration ../get started/configuration.md for detailed parameter configuration. 证据：`qwen-agent-docs/website/content/en/guide/core_moduls/tool.md`
- **Configuration**（documentation）：This document explains all configuration parameters of Agent. 证据：`qwen-agent-docs/website/content/en/guide/get_started/configuration.md`
- **Qwen-Agent Features**（documentation）：Qwen-Agent is a powerful and flexible framework for building intelligent LLM-powered applications. Key features include: 证据：`qwen-agent-docs/website/content/en/guide/get_started/features.md`
- **QuickStart**（documentation）：This quickstart will guide you to implement an agent using a few lines of code in just a few minutes. 证据：`qwen-agent-docs/website/content/en/guide/get_started/quickstart.md`
- **Qwen-Agent Overview**（documentation）：Qwen-Agent is a framework for developing LLM applications based on the instruction following, tool usage, planning, and memory capabilities of Qwen. 证据：`qwen-agent-docs/website/content/en/guide/index.md`
- **Qwen Agent Documentation**（documentation）：Welcome to the Qwen Agent documentation. Qwen Agent is an open-source AI agent framework that helps you build powerful LLM applications based on the instruction following, tool usage, planning, and memory capabilities of Qwen. 证据：`qwen-agent-docs/website/content/en/index.md`
- **Manifest**（structured_config）：{ "name": "BrowserQwen", "description" : "An Extension Driven by LLM", "version": "1.0", "manifest version": 3, 证据：`browser_qwen/manifest.json`
- **Server Config**（structured_config）：{ "path": { "work space root": "workspace/", "download root": "workspace/download/", "code interpreter ws": "workspace/tools/code interpreter/" }, "server": { "server host": "127.0.0.1", "fast api port": 7866, "app in browser port": 7863, "workstation port": 7864, "model server": "dashscope", "api key": "", "llm": "qwen-plus", "max ref token": 4000, "max days": 7 } } 证据：`qwen_server/server_config.json`
- **Models Config**（structured_config）：{ "models": { "qwen-plus": { "model name": "qwen-plus", "model type": "openai", "base url": "https://dashscope.aliyuncs.com/compatible-mode/v1", "api key env": "DASHSCOPE API KEY", "temperature": 0.0 }, "qwen3-max": { "model name": "qwen3-max", "model type": "openai", "base url": "https://dashscope.aliyuncs.com/compatible-mode/v1", "api key env": "DASHSCOPE API KEY", "temperature": 0.0 }, "gpt-4o-2024-11-20": { "model name": "gpt-4o-2024-11-20", "model type": "openai", "base url": "https://api.openai.com/v1/models", "api key env": "OPENAI API KEY", "temperature": 0.0 }, "gpt-5-2025-08-07-high": { "model name": "gpt-5-2025-08-07", "model type": "openai", "base url": "https://api.openai.com/v… 证据：`benchmark/deepplanning/models_config.json`
- **Level 1 Query Meta**（structured_config）：{ "id": "1", "query": "I'm putting together a complete footwear collection and need to order several specific items online. First, I'm looking for something from Nike in orange that has strong customer satisfaction - it needs fewer than 10 one-star reviews and more than 300 four-star reviews to ensure quality. Next, I need the Men's Puma RS-X Reinvention Classic White Sneakers from Puma, and since I need them quickly, the transport time must be less than 2 days. This item should have more than 3000 total reviews but fewer than 30 two-star reviews to confirm it's well-received. I also need an all-seasons product that can arrive within 1 day and has fewer than 30 two-star reviews for reliabil… 证据：`benchmark/deepplanning/shoppingplanning/data/level_1_query_meta.json`
- **Level 2 Query Meta**（structured_config）：{ "id": "1", "query": "I'm updating my wardrobe for an upcoming trip and need to order a few specific things. First, I'm looking for a 'Henley Top' from the brand Timberland that is popular, with monthly sales over 350 and fewer than 10 one-star reviews. Next, I need something from Ralph Lauren with an average rating score greater than 4.5 that can get here in less than 3 days. I also need a high-performance item for women from Arc'teryx that has to arrive in under 2 days; it must be a bestseller with more than 1900 total sales and over 300 five-star ratings.\n\nAdditionally, I'm searching for a black item with an average score above 4.5, monthly sales of more than 200, and fewer than 10 tw… 证据：`benchmark/deepplanning/shoppingplanning/data/level_2_query_meta.json`
- **Level 3 Query Meta**（structured_config）：{ "id": "1", "query": "I'm preparing for a weekend getaway and need to order several items with fast delivery. First, I need a highly-rated product that can arrive quickly—it must have an average rating above 4.5, more than 250 five-star reviews, fewer than 5 one-star reviews, and a transport time under 2 days. Next, I'm looking for Canvas Low-Top Sneakers in size 39 that are well-stocked, with inventory over 200 units to ensure availability. Then, I need something in size XL that's extremely popular and can get here even faster—it should have more than 300 five-star reviews, total sales exceeding 3000, and delivery within just 1 day. Finally, I want to grab Heritage Leather Ankle Boots tha… 证据：`benchmark/deepplanning/shoppingplanning/data/level_3_query_meta.json`
- **Shopping Tool Schema**（structured_config）：{ "type": "function", "function": { "name": "search products", "description": "Handles broad, open-ended natural language queries. It performs a semantic search on key product information name, brand, category, tags to quickly retrieve an initial set of relevant products. This is the first step in user exploration.", "parameters": { "type": "object", "properties": { "query": { "type": "string", "description": "User's natural language query. e.g., 'Uniqlo T-shirt', 'Nike running shoes'." }, "limit": { "type": "integer", "description": "Optional. Limits the number of product ids returned. Defaults to 20 to prevent excessively large results." } }, "required": "query" } } }, { "type": "function… 证据：`benchmark/deepplanning/shoppingplanning/tools/shopping_tool_schema.json`
- **Travelplanning Query En**（structured_config）：{ "id": "0", "query": "I'm planning a two-day trip from Hefei to Nanjing on November 12, 2025, returning in the evening of the 13th. The total budget for this trip should be within 3000 yuan. There are three of us traveling, and we'll take the train since it should be quite convenient—please help me choose a suitable train schedule.\n\nFor accommodation, I have specific preferences: I'd like a three-star hotel with a swimming pool, and please book two rooms.\n\nThere are several places I must visit during this trip, including 'Nanjing Deji Plaza' and 'Nanjing City Wall Taicheng Scenic Area'—please make sure to include both in the itinerary. Also, I’d like to have a meal near 'Laomendong', p… 证据：`benchmark/deepplanning/travelplanning/data/travelplanning_query_en.json`
- **Travelplanning Query Zh**（structured_config）：{ "id": "0", "query": "我打算2025年11月12号从合肥去南京玩两天，13号晚上就回来了，这次旅行的总开销希望控制在3000元以内。我们一共三个人，交通的话就坐火车吧，应该挺方便的，你帮我挑个合适的车次安排一下。住的地方我有点讲究，想找个带泳池的三星级酒店，顺便订两间房。\n\n这次南京行有几个地方是一定要去的，比如'南京德基广场'和'南京城墙台城景区'，这两个地方麻烦帮我都安排到行程里。另外啊，我想在'老门东'附近吃一顿饭，最好是能有生日套餐服务的餐厅，因为其中一位朋友正好生日，想一起庆祝一下。\n\n基本就是这些了，信息我都讲清楚了，你直接帮我规划一下行程吧！ 2025年11月12号是周三", "meta info": { "org": "合肥", "dest": "南京" , "days": 2, "depart date": "2025-11-12", "depart weekday": 3, "return date": "2025-11-13", "people number": 3, "hard constraints": { "train seat status": { "constraint context": "出行人数为3人，选择火车出行，需要选择剩余票数足够的列车（这句话不需要在query中体现，而是由用户自行推理得到）", "people number": 3, "outbound train no": "G7798", "inbound train no": "G3031", "outbound seat status":… 证据：`benchmark/deepplanning/travelplanning/data/travelplanning_query_zh.json`
- **Tool Schema**（structured_config）：{ "type": "function", "function": { "name": "query train info", "description": "查询火车票信息，支持按出发地、目的地、日期等条件查询火车票数据。返回符合条件的火车车次信息，包括：车次号、出发/到达时间、车站信息、行程时长、座位等级、余票状态、价格等。支持直达和中转方案，直达方案显示仅显示第一段行程，中转方案会显示各段行程的详细信息。在旅行计划中，所有火车相关信息必须严格来源于本工具返回的结果，不得自行编造；车站及列车相关名称必须与查询结果完全一致。", "parameters": { "type": "object", "properties": { "origin": { "type": "string", "description": "出发地，支持传城市名" }, "destination": { "type": "string", "description": "目的地，支持传城市名" }, "depDate": { "type": "string", "description": "出发日期，格式: YYYY-MM-DD" }, "seatClassName": { "type": "string", "description": "火车舱位等级：一等座,二等座,商务座" } }, "required": "origin", "destination", "depDate" } } }, { "type": "function", "function": { "name": "query… 证据：`benchmark/deepplanning/travelplanning/tools/tool_schema.json`
- **Tool Schema En**（structured_config）：{ "type": "function", "function": { "name": "query train info", "description": "Query train ticket information, supports searching train ticket data by origin, destination, date and other conditions. Returns train information that meets the criteria, including: train number, departure/arrival time, station information, journey duration, seat class, remaining seats status, price, etc. Supports direct and transfer routes. Direct routes display only the first segment, while transfer routes display detailed information for each segment. In travel planning, all train-related information must strictly come from the results returned by this tool and must not be fabricated; station and train names… 证据：`benchmark/deepplanning/travelplanning/tools/tool_schema_en.json`
- **Tool Schema Zh**（structured_config）：{ "type": "function", "function": { "name": "query train info", "description": "查询火车票信息，支持按出发地、目的地、日期等条件查询火车票数据。返回符合条件的火车车次信息，包括：车次号、出发/到达时间、车站信息、行程时长、座位等级、余票状态、价格等。支持直达和中转方案，直达方案显示仅显示第一段行程，中转方案会显示各段行程的详细信息。在旅行计划中，所有火车相关信息必须严格来源于本工具返回的结果，不得自行编造；车站及列车相关名称必须与查询结果完全一致。", "parameters": { "type": "object", "properties": { "origin": { "type": "string", "description": "出发地，支持传城市名" }, "destination": { "type": "string", "description": "目的地，支持传城市名" }, "depDate": { "type": "string", "description": "出发日期，格式: YYYY-MM-DD" }, "seatClassName": { "type": "string", "description": "火车舱位等级：一等座,二等座,商务座" } }, "required": "origin", "destination", "depDate" } } }, { "type": "function", "function": { "name": "query… 证据：`benchmark/deepplanning/travelplanning/tools/tool_schema_zh.json`
- **Tsconfig**（structured_config）：{ "compilerOptions": { "lib": "dom", "dom.iterable", "esnext" , "allowJs": true, "skipLibCheck": true, "strict": false, "noEmit": true, "incremental": true, "module": "esnext", "esModuleInterop": true, "moduleResolution": "bundler", "resolveJsonModule": true, "isolatedModules": true, "jsx": "preserve", "plugins": { "name": "next" } , "baseUrl": ".", "paths": { "@/ ": "./src/ " } }, "include": "next-env.d.ts", ".next/types/ / .ts", " / .ts", " / .tsx" , "exclude": "node modules" } 证据：`qwen-agent-docs/website/tsconfig.json`
- **Website Next.js/Node.js**（source_file）：.idea .vscode .DS Store .ipynb checkpoints 证据：`.gitignore`
- **.Pre Commit Config**（source_file）：repos: - repo: https://github.com/pycqa/flake8.git rev: 5.0.4 hooks: - id: flake8 args: "--max-line-length=300", "--extend-ignore=E231,E702,E251,W604" - repo: https://github.com/PyCQA/isort.git rev: 5.11.5 hooks: - id: isort args: "--line-length", "120" - repo: https://github.com/pre-commit/mirrors-yapf.git rev: v0.32.0 hooks: - id: yapf args: "--style", "{based on style: google, column limit: 120}", "-i" - repo: https://github.com/pre-commit/pre-commit-hooks.git rev: v4.3.0 hooks: - id: trailing-whitespace - id: check-yaml - id: end-of-file-fixer - id: requirements-txt-fixer - id: double-quote-string-fixer - id: check-merge-conflict - id: fix-encoding-pragma args: "--remove" - id: mixed-li… 证据：`.pre-commit-config.yaml`
- **Manifest**（source_file）：include qwen agent/utils/qwen.tiktoken recursive-include qwen agent/tools/resource 证据：`MANIFEST.in`
- **Background**（source_file）：function send data msg 证据：`browser_qwen/background.js`
- **Assistant Add Custom Tool**（source_file）：ROOT RESOURCE = os.path.join os.path.dirname file , 'resource' ⋮---- @register tool 'my image gen' class MyImageGen BaseTool ⋮---- description = 'AI painting image generation service, input text description, and return the image URL drawn based on text information.' parameters = { ⋮---- def call self, params: str, kwargs - str ⋮---- prompt = json5.loads params 'prompt' prompt = urllib.parse.quote prompt ⋮---- def init agent service ⋮---- llm cfg = {'model': 'qwen-max'} system = "According to the user's request, you first draw a picture and then automatically " ⋮---- tools = bot = Assistant ⋮---- def test query: str = 'draw a dog' ⋮---- bot = init agent service ⋮---- messages = {'role': 'use… 证据：`examples/assistant_add_custom_tool.py`
- **Assistant Audio**（source_file）：def test ⋮---- bot = Assistant llm={'model type': 'qwenaudio dashscope', 'model': 'qwen-audio-turbo-latest'} messages = { ⋮---- def app gui ⋮---- bot = Assistant llm={'model': 'qwen-audio-turbo-latest'} 证据：`examples/assistant_audio.py`
- **Assistant Mcp Sqlite Bot**（source_file）：ROOT RESOURCE = os.path.join os.path.dirname file , 'resource' ⋮---- def init agent service ⋮---- llm cfg = {'model': 'qwen-max'} system = '你扮演一个数据库助手，你具有查询数据库的能力' tools = { bot = Assistant ⋮---- def test query='数据库里有几张表', file: Optional str = os.path.join ROOT RESOURCE, 'poem.pdf' ⋮---- bot = init agent service ⋮---- messages = ⋮---- def app tui ⋮---- query = input 'user question: ' ⋮---- file = input 'file url press enter if no file : ' .strip ⋮---- response = ⋮---- def app gui ⋮---- chatbot config = { 证据：`examples/assistant_mcp_sqlite_bot.py`
- **Assistant Omni**（source_file）：def test ⋮---- bot = Assistant messages = { ⋮---- def app gui 证据：`examples/assistant_omni.py`
- **Define the agent**（source_file）：def init agent service ⋮---- llm cfg = { ⋮---- tools = { bot = Assistant ⋮---- def test query: str = 'What time is it?' ⋮---- bot = init agent service ⋮---- messages = {'role': 'user', 'content': query} response plain text = '' ⋮---- response plain text = typewriter print response, response plain text ⋮---- def app tui ⋮---- Define the agent ⋮---- Chat messages = ⋮---- query = input 'user question: ' ⋮---- response = ⋮---- def app gui ⋮---- chatbot config = { 证据：`examples/assistant_qwen3.5.py`
- **Define the agent**（source_file）：def init agent service ⋮---- llm cfg = { ⋮---- tools = bot = Assistant llm=llm cfg, ⋮---- def test query: str = 'What time is it?' ⋮---- bot = init agent service ⋮---- messages = {'role': 'user', 'content': query} response plain text = '' ⋮---- response plain text = typewriter print response, response plain text ⋮---- def app tui ⋮---- Define the agent ⋮---- Chat messages = ⋮---- query = input 'user question: ' ⋮---- response = ⋮---- def app gui ⋮---- chatbot config = { 证据：`examples/assistant_qwen3.py`
- **Define the agent**（source_file）：def init agent service ⋮---- llm cfg = { ⋮---- tools = bot = Assistant ⋮---- def test query: str = 'What time is it?' ⋮---- bot = init agent service ⋮---- messages = {'role': 'user', 'content': query} response plain text = '' ⋮---- response plain text = typewriter print response, response plain text ⋮---- def app tui ⋮---- Define the agent ⋮---- Chat messages = ⋮---- query = input 'user question: ' ⋮---- response = ⋮---- def app gui ⋮---- chatbot config = { 证据：`examples/assistant_qwen3_coder.py`
- **Define the agent**（source_file）：def init agent service ⋮---- llm cfg = { ⋮---- tools = bot = FnCallAgent ⋮---- def test pic url: str, query: str ⋮---- Define the agent bot = init agent service ⋮---- Chat messages = { ⋮---- response = list bot.run messages=messages -1 ⋮---- response plain text = response -1 'content' 证据：`examples/assistant_qwen3vl.py`
- **Define the agent**（source_file）：ROOT RESOURCE = os.path.join os.path.dirname file , 'resource' ⋮---- def init agent service ⋮---- llm cfg = { tools = bot = Assistant ⋮---- def test query: str = '画一只猫，再画一只狗，最后画他们一起玩的画面，给我三张图' ⋮---- bot = init agent service ⋮---- messages = {'role': 'user', 'content': query} response plain text = '' ⋮---- response plain text = typewriter print response, response plain text ⋮---- def app tui ⋮---- Define the agent ⋮---- Chat messages = ⋮---- query = input 'user question: ' ⋮---- response = ⋮---- def app gui ⋮---- chatbot config = {'prompt.suggestions': '画一只猫，再画一只狗，最后画他们一起玩的画面，给我三张图' } 证据：`examples/assistant_qwq.py`
- **Assistant Rag**（source_file）：def test ⋮---- bot = Assistant llm={'model': 'qwen-plus-latest'} messages = {'role': 'user', 'content': {'text': '介绍图一'}, {'file': 'https://arxiv.org/pdf/1706.03762.pdf'} } ⋮---- def app gui ⋮---- bot = Assistant llm={'model': 'qwen-plus-latest'}, chatbot config = { 证据：`examples/assistant_rag.py`
- **Assistant Weather Bot**（source_file）：ROOT RESOURCE = os.path.join os.path.dirname file , 'resource' ⋮---- def init agent service ⋮---- llm cfg = {'model': 'qwen-max'} system = '你扮演一个天气预报助手，你具有查询天气和画图能力。' ⋮---- tools = 'image gen', 'amap weather' bot = Assistant ⋮---- def test query='海淀区天气', file: Optional str = os.path.join ROOT RESOURCE, 'poem.pdf' ⋮---- bot = init agent service ⋮---- messages = ⋮---- def app tui ⋮---- query = input 'user question: ' ⋮---- file = input 'file url press enter if no file : ' .strip ⋮---- response = ⋮---- def app gui ⋮---- chatbot config = { 证据：`examples/assistant_weather_bot.py`
- **Function Calling**（source_file）：def get current weather location, unit='fahrenheit' ⋮---- def test fncall prompt type: str = 'qwen' ⋮---- llm = get chat model { ⋮---- messages = {'role': 'user', 'content': "What's the weather like in San Francisco?"} functions = { ⋮---- responses = ⋮---- last response = messages -1 ⋮---- available functions = { function name = last response 'function call' 'name' function to call = available functions function name function args = json.loads last response 'function call' 'arguments' function response = function to call 证据：`examples/function_calling.py`
- **Function Calling In Parallel**（source_file）：def get current weather location, unit='fahrenheit' ⋮---- def test ⋮---- llm = get chat model { ⋮---- messages = { functions = { ⋮---- responses = ⋮---- fncall msgs = rsp for rsp in responses if rsp.get 'function call', None ⋮---- available functions = { ⋮---- function name = msg 'function call' 'name' function to call = available functions function name function args = json.loads msg 'function call' 'arguments' function response = function to call 证据：`examples/function_calling_in_parallel.py`
- **Gpt Mentions**（source_file）：def init agent service ⋮---- llm cfg = {'model': 'qwen-max'} ⋮---- react chat agent = ReActChat doc qa agent = BasicDocQA ⋮---- assistant agent = Assistant llm=llm cfg, name='小助理', description="I'm a helpful assistant" ⋮---- def app gui ⋮---- agent list = init agent service chatbotConfig = { 证据：`examples/gpt_mentions.py`
- **Group Chat Chess**（source_file）：NPC NAME = '小明' USER NAME = '小塘' CFGS = { ⋮---- def test query: str = ' ' ⋮---- bot = GroupChat agents=CFGS, llm={'model': 'qwen-max'} ⋮---- messages = Message 'user', query, name=USER NAME ⋮---- def app tui ⋮---- messages = ⋮---- query = input 'user question: ' ⋮---- response = ⋮---- def app gui ⋮---- chatbot config = { 证据：`examples/group_chat_chess.py`
- **The topic ends**（source_file）：def init agent service cfgs ⋮---- llm cfg = {'model': 'qwen-max'} bot = GroupChat agents=cfgs, llm=llm cfg ⋮---- def init agent service create ⋮---- bot = GroupChatCreator llm=llm cfg ⋮---- app global para = { ⋮---- CFGS = { ⋮---- MAX ROUND = 3 ⋮---- def app cfgs ⋮---- cfgs = json5.loads cfgs bot = init agent service cfgs=cfgs ⋮---- mentioned agents name = ⋮---- messages = app global para 'messages' ⋮---- content = '' ⋮---- content = '\n'.join x.text if x.text else '' for x in messages -1 .content .strip ⋮---- content = messages -1 .content.strip ⋮---- response = ⋮---- display history = get display history from message ⋮---- incremental history = ⋮---- function display = '' ⋮---- function d… 证据：`examples/group_chat_demo.py`
- **Llm Quick Chat Oai**（source_file）：def test ⋮---- llm cfg = {'model': 'qwen-max', 'model server': 'dashscope'} tools = { ⋮---- llm = get chat model llm cfg messages = {'role': 'user', 'content': '你是？'} ⋮---- response = llm.quick chat oai messages ⋮---- response = llm.quick chat oai messages, tools=tools 证据：`examples/llm_quick_chat_oai.py`
- **Llm Riddles**（source_file）：class LLMRiddles Agent ⋮---- def init self, llm: Optional Union Dict, BaseChatModel = None ⋮---- def run self, messages: List Message , lang: str = 'en', kwargs - Iterator List Message ⋮---- def test ⋮---- bot = LLMRiddles llm={'model': 'qwen-max'} ⋮---- messages = query = f'请直接输出“{topic}”（不需要引号），不要说其他内容' ⋮---- response = ⋮---- def app tui ⋮---- query = input 'user question input EXIT for next topic : ' 证据：`examples/llm_riddles.py`
- **Llm Vl Mix Text**（source_file）：def test ⋮---- llm cfg = {'model': 'qwen-max', 'model server': 'dashscope'} llm cfg vl = {'model': 'qwen-vl-max', 'model server': 'dashscope'} functions = { ⋮---- llm vl = get chat model llm cfg vl messages = { response = llm vl.chat messages, stream=True ⋮---- llm = get chat model llm cfg ⋮---- response = llm.chat messages, stream=True ⋮---- response = llm.chat messages, functions=functions, stream=True 证据：`examples/llm_vl_mix_text.py`
- 其余 17 条证据见 `AI_CONTEXT_PACK.json` 或 `EVIDENCE_INDEX.json`。

## 宿主 AI 必须遵守的规则

- **把本资产当作开工前上下文，而不是运行环境。**：AI Context Pack 只包含证据化项目理解，不包含目标项目的可执行状态。 证据：`README.md`, `benchmark/code_interpreter/README.md`, `benchmark/deepplanning/README.md`
- **回答用户时区分可预览内容与必须安装后才能验证的内容。**：安装前体验的消费者价值来自降低误装和误判，而不是伪装成真实运行。 证据：`README.md`, `benchmark/code_interpreter/README.md`, `benchmark/deepplanning/README.md`

## 用户开工前应该回答的问题

- 你准备在哪个宿主 AI 或本地环境中使用它？
- 你只是想先体验工作流，还是准备真实安装？
- 你最在意的是安装成本、输出质量、还是和现有规则的冲突？

## 验收标准

- 所有能力声明都能回指到 evidence_refs 中的文件路径。
- AI_CONTEXT_PACK.md 没有把预览包装成真实运行。
- 用户能在 3 分钟内看懂适合谁、能做什么、如何开始和风险边界。

---

## Doramagic Context Augmentation

下面内容用于强化 Repomix/AI Context Pack 主体。Human Manual 只提供阅读骨架；踩坑日志会被转成宿主 AI 必须遵守的工作约束。

## Human Manual 骨架

使用规则：这里只是项目阅读路线和显著性信号，不是事实权威。具体事实仍必须回到 repo evidence / Claim Graph。

宿主 AI 硬性规则：
- 不得把页标题、章节顺序、摘要或 importance 当作项目事实证据。
- 解释 Human Manual 骨架时，必须明确说它只是阅读路线/显著性信号。
- 能力、安装、兼容性、运行状态和风险判断必须引用 repo evidence、source path 或 Claim Graph。

- **Qwen-Agent 简介**：importance `high`
  - source_paths: README.md, qwen_agent/__init__.py
- **快速开始**：importance `high`
  - source_paths: examples/assistant_add_custom_tool.py, qwen_agent/agents/assistant.py
- **安装指南**：importance `high`
  - source_paths: setup.py, qwen_agent/tools/code_interpreter.py, qwen_agent/tools/resource/code_interpreter_image.dockerfile
- **核心模块架构**：importance `high`
  - source_paths: qwen_agent/agent.py, qwen_agent/llm/base.py, qwen_agent/tools/base.py, qwen_agent/memory/memory.py
- **Agent 系统详解**：importance `high`
  - source_paths: qwen_agent/agents/assistant.py, qwen_agent/agents/fncall_agent.py, qwen_agent/agents/react_chat.py, qwen_agent/agents/group_chat.py, qwen_agent/agents/__init__.py
- **大模型集成**：importance `high`
  - source_paths: qwen_agent/llm/qwen_dashscope.py, qwen_agent/llm/oai.py, qwen_agent/llm/function_calling.py, qwen_agent/llm/fncall_prompts/base_fncall_prompt.py, examples/function_calling.py
- **工具使用指南**：importance `high`
  - source_paths: qwen_agent/tools/base.py, qwen_agent/tools/doc_parser.py, qwen_agent/tools/image_gen.py, qwen_agent/tools/web_search.py, qwen_agent/tools/retrieval.py
- **代码解释器**：importance `high`
  - source_paths: qwen_agent/tools/code_interpreter.py, qwen_agent/tools/python_executor.py, qwen_agent/tools/resource/code_interpreter_image.dockerfile, qwen_agent/tools/resource/code_interpreter_init_kernel.py, examples/tir_math.py

## Repo Inspection Evidence / 源码检查证据

- repo_clone_verified: true
- repo_inspection_verified: true
- repo_commit: `31a4d36d123688581a9e9744427272b33ce940e0`
- inspected_files: `README.md`, `examples/llm_quick_chat_oai.py`, `examples/function_calling_in_parallel.py`, `examples/assistant_qwq.py`, `examples/function_calling.py`, `examples/assistant_rag.py`, `examples/multi_agent_router.py`, `examples/visual_storytelling.py`, `examples/assistant_omni.py`, `examples/assistant_weather_bot.py`, `examples/llm_riddles.py`, `examples/assistant_qwen3.5.py`, `examples/gpt_mentions.py`, `examples/assistant_qwen3vl.py`, `examples/qwen2vl_function_calling.py`, `examples/qwen2vl_assistant_tooluse.py`, `examples/assistant_qwen3.py`, `examples/tir_math.py`, `examples/assistant_add_custom_tool.py`, `examples/group_chat_chess.py`

宿主 AI 硬性规则：
- 没有 repo_clone_verified=true 时，不得声称已经读过源码。
- 没有 repo_inspection_verified=true 时，不得把 README/docs/package 文件判断写成事实。
- 没有 quick_start_verified=true 时，不得声称 Quick Start 已跑通。

## Doramagic Pitfall Constraints / 踩坑约束

这些规则来自 Doramagic 发现、验证或编译过程中的项目专属坑点。宿主 AI 必须把它们当作工作约束，而不是普通说明文字。

### Constraint 1: 来源证据：Security: Unsandboxed exec()/eval() in PythonExecutor with trivially bypassable filter

- Trigger: GitHub 社区证据显示该项目存在一个安全/权限相关的待验证问题：Security: Unsandboxed exec()/eval() in PythonExecutor with trivially bypassable filter
- Host AI rule: 来源问题仍为 open，Pack Agent 需要复核是否仍影响当前版本。
- Why it matters: 可能增加新用户试用和生产接入成本。
- Evidence: community_evidence:github | cevd_7d5e82403e6949e0bd07e5969ad25506 | https://github.com/QwenLM/Qwen-Agent/issues/866 | 来源讨论提到 python 相关条件，需在安装/试用前复核。
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 2: 来源证据：coder.qwen.ai "Publish to GitHub" fail,s

- Trigger: GitHub 社区证据显示该项目存在一个安全/权限相关的待验证问题：coder.qwen.ai "Publish to GitHub" fail,s
- Host AI rule: 来源问题仍为 open，Pack Agent 需要复核是否仍影响当前版本。
- Why it matters: 可能影响授权、密钥配置或安全边界。
- Evidence: community_evidence:github | cevd_c33922a5c96145f790509151aad00dab | https://github.com/QwenLM/Qwen-Agent/issues/783 | 来源类型 github_issue 暴露的待验证使用条件。
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 3: 能力判断依赖假设

- Trigger: README/documentation is current enough for a first validation pass.
- Host AI rule: 将假设转成下游验证清单。
- Why it matters: 假设不成立时，用户拿不到承诺的能力。
- Evidence: capability.assumptions | art_c108f7398a1c4fdc998a669640a849ed | https://github.com/QwenLM/Qwen-Agent#readme | README/documentation is current enough for a first validation pass.
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 4: 来源证据：Qwen3.5是否支持image_zoom_in_tool

- Trigger: GitHub 社区证据显示该项目存在一个运行相关的待验证问题：Qwen3.5是否支持image_zoom_in_tool
- Host AI rule: 来源问题仍为 open，Pack Agent 需要复核是否仍影响当前版本。
- Why it matters: 可能增加新用户试用和生产接入成本。
- Evidence: community_evidence:github | cevd_11290573df78457d94c7237d599c3cf5 | https://github.com/QwenLM/Qwen-Agent/issues/879 | 来源类型 github_issue 暴露的待验证使用条件。
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 5: 维护活跃度未知

- Trigger: 未记录 last_activity_observed。
- Host AI rule: 补 GitHub 最近 commit、release、issue/PR 响应信号。
- Why it matters: 新项目、停更项目和活跃项目会被混在一起，推荐信任度下降。
- Evidence: evidence.maintainer_signals | art_c108f7398a1c4fdc998a669640a849ed | https://github.com/QwenLM/Qwen-Agent#readme | last_activity_observed missing
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 6: 下游验证发现风险项

- Trigger: no_demo
- Host AI rule: 进入安全/权限治理复核队列。
- Why it matters: 下游已经要求复核，不能在页面中弱化。
- Evidence: downstream_validation.risk_items | art_c108f7398a1c4fdc998a669640a849ed | https://github.com/QwenLM/Qwen-Agent#readme | no_demo; severity=medium
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 7: 存在评分风险

- Trigger: no_demo
- Host AI rule: 把风险写入边界卡，并确认是否需要人工复核。
- Why it matters: 风险会影响是否适合普通用户安装。
- Evidence: risks.scoring_risks | art_c108f7398a1c4fdc998a669640a849ed | https://github.com/QwenLM/Qwen-Agent#readme | no_demo; severity=medium
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 8: issue/PR 响应质量未知

- Trigger: issue_or_pr_quality=unknown。
- Host AI rule: 抽样最近 issue/PR，判断是否长期无人处理。
- Why it matters: 用户无法判断遇到问题后是否有人维护。
- Evidence: evidence.maintainer_signals | art_c108f7398a1c4fdc998a669640a849ed | https://github.com/QwenLM/Qwen-Agent#readme | issue_or_pr_quality=unknown
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 9: 发布节奏不明确

- Trigger: release_recency=unknown。
- Host AI rule: 确认最近 release/tag 和 README 安装命令是否一致。
- Why it matters: 安装命令和文档可能落后于代码，用户踩坑概率升高。
- Evidence: evidence.maintainer_signals | art_c108f7398a1c4fdc998a669640a849ed | https://github.com/QwenLM/Qwen-Agent#readme | release_recency=unknown
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。