# javascript-sdk - 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 javascript-sdk 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_0019` supported 0.86

## 它能做什么

- **Natural Language Web Querying**（需要安装后验证）：AgentQL uses a natural language query language to find elements and extract data from web pages, with self-healing behavior when UI changes. 证据：`README.md`, `README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Python SDK Web Automation**（需要安装后验证）：Python SDK that integrates with Playwright to execute AgentQL queries for web scraping and automation tasks. 证据：`README.md`, `examples/python/first_steps/README.md`, `examples/python/package.json (referenced via examples)` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **JavaScript SDK Web Automation**（需要安装后验证）：JavaScript/Node.js SDK for executing AgentQL queries, compatible with Cloudflare Browser Rendering. 证据：`README.md`, `examples/js/first-steps/README.md`, `examples/js/package.json` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Structured Data Extraction**（需要安装后验证）：Extract structured data from web pages by defining the shape of output in the AgentQL query. 证据：`README.md`, `examples/python/collect_ecommerce_pricing_data/README.md`, `examples/js/collect-pricing-data/README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Bot Detection Evasion (Stealth Mode)**（需要安装后验证）：Techniques to avoid anti-bot detection including randomized HTTP headers, browser fingerprinting, timezone/geolocation randomization, and humanlike mouse movement. 证据：`examples/js/stealth-mode/README.md`, `examples/python/stealth_mode/README.md`, `examples/js/humanlike-antibot/README.md`, `examples/python/humanlike-antibot/README.md` Claim：`clm_0005` supported 0.86
- **Cross-Site Compatible Queries**（需要安装后验证）：Use the same AgentQL query across different websites with similar content structures. 证据：`README.md`, `examples/js/compare-product-prices/README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Authenticated Session Persistence**（需要安装后验证）：Save and restore browser authenticated state (signed-in sessions) to maintain login across automation runs. 证据：`examples/js/save-and-load-authenticated-session/README.md`, `examples/python/save_and_load_authenticated_session/README.md` Claim：`clm_0007` supported 0.86
- **Pagination and Infinite Scroll Handling**（需要安装后验证）：Navigate through paginated content and handle infinite scroll pages to collect data across multiple pages. 证据：`examples/js/collect-paginated-ecommerce-data/README.md`, `examples/js/infinite-scroll/README.md`, `examples/python/infinite_scroll/README.md`, `examples/python/collect_paginated_news_headlines/README.md` Claim：`clm_0008` supported 0.86
- **Remote Browser Support**（需要安装后验证）：Connect to remotely hosted browsers via Chrome DevTools Protocol (CDP) for cloud-based or distributed automation. 证据：`examples/js/use-remote-browser/README.md`, `examples/python/use_remote_browser/README.md` Claim：`clm_0009` supported 0.86
- **Existing Browser Connection**（需要安装后验证）：Attach to user's existing Chrome browser via remote debugging port for automation using personal profiles and sessions. 证据：`examples/js/use-existing-browser/README.md`, `examples/python/use_existing_browser/README.md` Claim：`clm_0010` supported 0.86
- **Headless Browser Execution**（需要安装后验证）：Run AgentQL scripts in headless browser mode for server-side automation without visible UI. 证据：`examples/js/run-script-in-headless-browser/README.md`, `examples/python/run_script_in_headless_browser/README.md` Claim：`clm_0011` supported 0.86
- **Chrome Debugger Extension**（可做安装前预览）：Browser extension for debugging and testing AgentQL queries in real-time on live websites. 证据：`README.md`, `examples/js/first-steps/README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **REST API Endpoint**（可做安装前预览）：Execute AgentQL queries via REST API without requiring an SDK installation. 证据：`README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Form Interaction Automation**（需要安装后验证）：Automate form filling, submission, and login processes using natural language element queries. 证据：`examples/js/submit-form/README.md`, `examples/js/log-into-sites/README.md`, `examples/python/submit_form/README.md`, `examples/python/log_into_sites/README.md` Claim：`clm_0014` supported 0.86
- **Google Colab Integration**（需要安装后验证）：Run AgentQL scripts in Google Colaboratory environment with video recording of browser sessions. 证据：`examples/python/run_script_online_in_google_colab/README.md`, `examples/googlecolab/close_cookie_dialog/main.ipynb` Claim：`clm_0015` supported 0.86
- **XPath Element Retrieval**（需要安装后验证）：Get XPath selectors for elements fetched with AgentQL queries for use with standard Playwright locators. 证据：`examples/js/xpath/README.md`, `examples/python/xpath/README.md` Claim：`clm_0016` supported 0.86
- **Element getByPrompt Method**（需要安装后验证）：Retrieve a single element using natural language prompt descriptions via get_by_prompt/getByPrompt method. 证据：`examples/js/get-by-prompt/README.md`, `examples/python/get_by_prompt/README.md` Claim：`clm_0017` supported 0.86
- **Agent Framework Integrations**（需要安装后验证）：Integrations with agent frameworks like Langchain and automation tools like Zapier via MCP server. 证据：`README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等

## 怎么开始

- `npm install openai` 证据：`examples/js/perform-sentiment-analysis/README.md` Claim：`clm_0020` supported 0.86
- `npm install playwright-dompath` 证据：`examples/js/xpath/README.md` Claim：`clm_0021` supported 0.86
- `pip install openai` 证据：`examples/python/perform_sentiment_analysis/README.md` Claim：`clm_0022` supported 0.86
- `pip install playwright-dompath` 证据：`examples/python/xpath/README.md` Claim：`clm_0023` supported 0.86

## 继续前判断卡

- **当前建议**：先做角色匹配试用
- **为什么**：这个项目更像角色库，核心风险是选错角色或把角色文案当执行能力；先用 Prompt Preview 试角色匹配，再决定是否沙盒导入。

### 30 秒判断

- **现在怎么做**：先做角色匹配试用
- **最小安全下一步**：先用 Prompt Preview 试角色匹配；满意后再隔离导入
- **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_0019` supported 0.86
- **能力存在：Natural Language Web Querying**（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`, `README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86
- **能力存在：Python SDK Web Automation**（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`, `examples/python/first_steps/README.md`, `examples/python/package.json (referenced via examples)` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86
- **能力存在：JavaScript SDK Web Automation**（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`, `examples/js/first-steps/README.md`, `examples/js/package.json` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86
- **能力存在：Structured Data Extraction**（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`, `examples/python/collect_ecommerce_pricing_data/README.md`, `examples/js/collect-pricing-data/README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86
- **能力存在：Bot Detection Evasion (Stealth Mode)**（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：`examples/js/stealth-mode/README.md`, `examples/python/stealth_mode/README.md`, `examples/js/humanlike-antibot/README.md`, `examples/python/humanlike-antibot/README.md` Claim：`clm_0005` 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：`examples/js/perform-sentiment-analysis/README.md`, `examples/js/xpath/README.md`, `examples/python/perform_sentiment_analysis/README.md`, `examples/python/xpath/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`, `examples/googlecolab/close_cookie_dialog/main.ipynb`, `examples/js/collect-paginated-ecommerce-data/README.md`, `examples/js/collect-pricing-data/README.md` 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。）
- **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。
- **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_0024` 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：`examples/js/perform-sentiment-analysis/README.md`, `examples/js/xpath/README.md`, `examples/python/perform_sentiment_analysis/README.md`, `examples/python/xpath/README.md` Claim：`clm_0025` supported 0.86
- **风险**： 处理方式：
- **风险**： 处理方式：
- **风险**： 处理方式：
- **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

- **Natural Language Web Querying**：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`, `README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Python SDK Web Automation**：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`, `examples/python/first_steps/README.md`, `examples/python/package.json (referenced via examples)` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **JavaScript SDK Web Automation**：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`, `examples/js/first-steps/README.md`, `examples/js/package.json` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Structured Data Extraction**：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`, `examples/python/collect_ecommerce_pricing_data/README.md`, `examples/js/collect-pricing-data/README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Bot Detection Evasion (Stealth Mode)**：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：`examples/js/stealth-mode/README.md`, `examples/python/stealth_mode/README.md`, `examples/js/humanlike-antibot/README.md`, `examples/python/humanlike-antibot/README.md` Claim：`clm_0005` supported 0.86
- **Cross-Site Compatible Queries**：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`, `examples/js/compare-product-prices/README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Authenticated Session Persistence**：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：`examples/js/save-and-load-authenticated-session/README.md`, `examples/python/save_and_load_authenticated_session/README.md` Claim：`clm_0007` supported 0.86
- **Pagination and Infinite Scroll Handling**：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：`examples/js/collect-paginated-ecommerce-data/README.md`, `examples/js/infinite-scroll/README.md`, `examples/python/infinite_scroll/README.md`, `examples/python/collect_paginated_news_headlines/README.md` Claim：`clm_0008` supported 0.86
- **Remote Browser Support**：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：`examples/js/use-remote-browser/README.md`, `examples/python/use_remote_browser/README.md` Claim：`clm_0009` supported 0.86
- **Existing Browser Connection**：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：`examples/js/use-existing-browser/README.md`, `examples/python/use_existing_browser/README.md` Claim：`clm_0010` supported 0.86
- **Headless Browser Execution**：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：`examples/js/run-script-in-headless-browser/README.md`, `examples/python/run_script_in_headless_browser/README.md` Claim：`clm_0011` supported 0.86
- **Chrome Debugger Extension**：先基于 role_skill_index / evidence_index 帮用户挑选可用角色、Skill 或工作流。 边界：可做安装前 Prompt 体验。 证据：`README.md`, `examples/js/first-steps/README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **REST API Endpoint**：先基于 role_skill_index / evidence_index 帮用户挑选可用角色、Skill 或工作流。 边界：可做安装前 Prompt 体验。 证据：`README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Form Interaction Automation**：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：`examples/js/submit-form/README.md`, `examples/js/log-into-sites/README.md`, `examples/python/submit_form/README.md`, `examples/python/log_into_sites/README.md` Claim：`clm_0014` supported 0.86
- **Google Colab Integration**：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：`examples/python/run_script_online_in_google_colab/README.md`, `examples/googlecolab/close_cookie_dialog/main.ipynb` Claim：`clm_0015` supported 0.86
- **XPath Element Retrieval**：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：`examples/js/xpath/README.md`, `examples/python/xpath/README.md` Claim：`clm_0016` supported 0.86
- **Element getByPrompt Method**：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：`examples/js/get-by-prompt/README.md`, `examples/python/get_by_prompt/README.md` Claim：`clm_0017` supported 0.86
- **Agent Framework Integrations**：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, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等

### 上下文规模

- 文件总数：133
- 重要文件覆盖：40/133
- 证据索引条目：76
- 角色 / Skill 条目：51

### 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 javascript-sdk 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 javascript-sdk 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 javascript-sdk 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 51 个角色 / Skill / 项目文档条目。

- **What is AgentQL?**（project_doc）：AgentQL connects LLMs and AI agents to the entire web 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`
- **Example script: title of example**（project_doc）：This is an example shows how to do a thing with AgentQL . 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：`.templates/python/README.md`
- **Example script: close a cookie dialog with AgentQL**（project_doc）：Close a site's dialog for rejecting/accepting cookies with AgentQL. 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：`examples/js/close-cookie-dialog/README.md`
- **Example script: closing popup windows with AgentQL**（project_doc）：Close modals like promotional forms and banners with AgentQL. 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：`examples/js/close-popup/README.md`
- **Example script: Collect ecommerce data across multiple paginated webpages**（project_doc）：Collect paginated bookstore listing data with AgentQL. 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：`examples/js/collect-paginated-ecommerce-data/README.md`
- **Example script: Collect news headlines across multiple paginated webpages**（project_doc）：Collect paginated data from HackerNews with AgentQL. 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：`examples/js/collect-paginated-news-headlines/README.md`
- **Example script: collecting pricing data from e-commerce website using AgentQL**（project_doc）：Collect pricing data from an e-commerce website using AgentQL. 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：`examples/js/collect-pricing-data/README.md`
- **Example script: interacting with YouTube website and extracting video information using AgentQL**（project_doc）：Interact with YouTube website and extract video information using AgentQL. 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：`examples/js/collect-youtube-comments/README.md`
- **Example script: comparing product price across websites with AgentQL**（project_doc）：Compare product prices across websites with query data method with AgentQL. 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：`examples/js/compare-product-prices/README.md`
- **Example script: collecting pricing data from e-commerce website using AgentQL**（project_doc）：Collect pricing data from an e-commerce website using AgentQL. 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：`examples/js/first-steps/README.md`
- **Example script: interact with elements leveraging getByPrompt method**（project_doc）：Use AgentQL's get by prompt method to retrieve an element and interact with it. 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：`examples/js/get-by-prompt/README.md`
- **Example script: use humanlike mouse movement to avoid antibot**（project_doc）：Use humanlike mouse movement to avoid antibot when using AgentQL. 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：`examples/js/humanlike-antibot/README.md`
- **Example script: load additional content on page by scrolling**（project_doc）：How to load additional content on pages that load content based on scroll position aka 'infinite scroll' . 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：`examples/js/infinite-scroll/README.md`
- **Example script: querying a list of items with AgentQL**（project_doc）：How to query a list of items on the page with AgentQL. 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：`examples/js/list-query-usage/README.md`
- **Example script: login into websites**（project_doc）：Use AgentQL to interact with login forms and log into websites. 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：`examples/js/log-into-sites/README.md`
- **Example script: querying a list of items with AgentQL**（project_doc）：How to query a list of items on the page with AgentQL. 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：`examples/js/maps_scraper/README.md`
- **Example script: perform sentiment analysis with AgentQL**（project_doc）：Perform sentiment analysis on YouTube comments with AgentQL and OpenAI's GPT-3.5 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：`examples/js/news-aggregator/README.md`
- **Example script: perform sentiment analysis with AgentQL**（project_doc）：Perform sentiment analysis on YouTube comments with AgentQL and OpenAI's GPT-3.5 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：`examples/js/perform-sentiment-analysis/README.md`
- **Example script: run the script in headless browser with AgentQL**（project_doc）：Run the script in headless browser with AgentQL. 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：`examples/js/run-script-in-headless-browser/README.md`
- **Example script: save and load authenticated state with AgentQL**（project_doc）：Save and load an authenticated state i.e. signed-in state with AgentQL. 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：`examples/js/save-and-load-authenticated-session/README.md`
- **Stealth mode: Running AgentQL in stealth mode and avoiding bot detection**（project_doc）：How to apply different techniques to lower the risk of being detected by an anti-bot system. 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：`examples/js/stealth-mode/README.md`
- **Example script: Automating form submission using AgentQL**（project_doc）：Submit form with AgentQL. 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：`examples/js/submit-form/README.md`
- **Example script: Use existing browser with AgentQL**（project_doc）：If you don't want to use Playwright's default browser, you can bring your own. 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：`examples/js/use-existing-browser/README.md`
- **Example script: Use remote browser with AgentQL**（project_doc）：Browser is hosted remotely - you connect to it over CDP. 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：`examples/js/use-remote-browser/README.md`
- **Example script: wait for page to load with AgentQL**（project_doc）：Wait for the page to load completely before querying the page with AgentQL. 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：`examples/js/wait-for-entire-page-load/README.md`
- **Example script: get XPath of elements fetched with AgentQL**（project_doc）：Get XPath of an element that was fetched with AgentQL. 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：`examples/js/xpath/README.md`
- **Example script: close a cookie dialog with AgentQL**（project_doc）：Close a site's dialog for rejecting/accepting cookies with AgentQL. 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：`examples/python/close_cookie_dialog/README.md`
- **Example script: closing popup windows with AgentQL**（project_doc）：Close popup windows like promotion forms with AgentQL. 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：`examples/python/close_popup/README.md`
- **Example script: collecting pricing data with AgentQL from e-commerce websites**（project_doc）：Collect pricing data from e-commerce websites using AgentQL. 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：`examples/python/collect_ecommerce_pricing_data/README.md`
- **Example script: Collect ecommerce data across multiple paginated webpages**（project_doc）：Collect ecommerce data across multiple paginated webpages. 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：`examples/python/collect_paginated_ecommerce_listing_data/README.md`
- **Example script: Collect news headlines across multiple paginated webpages**（project_doc）：Collect news headlines across multiple paginated webpages. 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：`examples/python/collect_paginated_news_headlines/README.md`
- **Example script: comparing product price across websites with AgentQL**（project_doc）：Compare product prices across websites with AgentQL. 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：`examples/python/compare_product_prices/README.md`
- **Example script: collecting pricing data from e-commerce website using AgentQL**（project_doc）：Collect pricing data from e-commerce website using AgentQL. 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：`examples/python/first_steps/README.md`
- **Example script: interact with elements leveraging get by prompt method**（project_doc）：Use AgentQL's get by prompt method to retrieve an element and interact with it. 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：`examples/python/get_by_prompt/README.md`
- **Example script: use humanlike mouse movement to avoid antibot**（project_doc）：Use humanlike mouse movement to avoid antibot with AgentQL. 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：`examples/python/humanlike-antibot/README.md`
- **Example script: load additional content on page by scrolling**（project_doc）：Load additional content on pages that load content based on scroll position. 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：`examples/python/infinite_scroll/README.md`
- **Example script: querying a list of items with AgentQL**（project_doc）：Query a list of items on the page with AgentQL. 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：`examples/python/list_query_usage/README.md`
- **Example script: logging into sites with AgentQL**（project_doc）：Use AgentQL to interact with login forms and log into websites. 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：`examples/python/log_into_sites/README.md`
- **Example script: scraping Google Maps with AgentQL**（project_doc）：Scrape data from maps using AgentQL. 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：`examples/python/maps_scraper/README.md`
- **Example script: News Aggregator to CSV**（project_doc）：Return news headlines from a selection of URLs and save the results to a CSV file. 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：`examples/python/news-aggregator/README.md`
- **Example script: performing sentiment analysis with AgentQL**（project_doc）：Perform sentiment analysis on YouTube comments with AgentQL and OpenAI's GPT-3.5 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：`examples/python/perform_sentiment_analysis/README.md`
- **Example script: running a script in headless browser with AgentQL**（project_doc）：Run the script in a headless browser with AgentQL. 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：`examples/python/run_script_in_headless_browser/README.md`
- **Example script: run AgentQL script in online development environment**（project_doc）：Run AgentQL script in online development environment like Google Colaboratory. 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：`examples/python/run_script_online_in_google_colab/README.md`
- **Example script: saving and loading authenticated state with AgentQL**（project_doc）：Save and load an authenticated state i.e., signed-in state with AgentQL. 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：`examples/python/save_and_load_authenticated_session/README.md`
- **Stealth mode: Running AgentQL in stealth mode and avoiding bot detection**（project_doc）：Apply techniques to lower the risk of being detected by anti-bot systems. 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：`examples/python/stealth_mode/README.md`
- **Example script: Filling out and submitting a form using AgentQL**（project_doc）：Submit form with AgentQL. 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：`examples/python/submit_form/README.md`
- **Example script: Use existing browser with AgentQL**（project_doc）：Use existing browser with AgentQL. 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：`examples/python/use_existing_browser/README.md`
- **Example script: Use remote browser with AgentQL**（project_doc）：Browser is hosted remotely - you connect to it over CDP. 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：`examples/python/use_remote_browser/README.md`
- **Example script: waiting for page to load with AgentQL**（project_doc）：Wait for the page to load completely before querying the page with AgentQL. 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：`examples/python/wait_for_entire_page_load/README.md`
- **Example script: getting an element's XPath with AgentQL**（project_doc）：Get XPath of an element that was fetched with AgentQL. 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：`examples/python/xpath/README.md`
- **Overview**（project_doc）：Browse our opensource collection of useful example scripts and get started extracting data, scraping, and automating. 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：`examples/examples.md`

## 证据索引

- 共索引 76 条证据。

- **What is AgentQL?**（documentation）：AgentQL connects LLMs and AI agents to the entire web 证据：`README.md`
- **Example script: title of example**（documentation）：This is an example shows how to do a thing with AgentQL . 证据：`.templates/python/README.md`
- **Example script: close a cookie dialog with AgentQL**（documentation）：Example script: close a cookie dialog with AgentQL 证据：`examples/js/close-cookie-dialog/README.md`
- **Example script: closing popup windows with AgentQL**（documentation）：Example script: closing popup windows with AgentQL 证据：`examples/js/close-popup/README.md`
- **Example script: Collect ecommerce data across multiple paginated webpages**（documentation）：Example script: Collect ecommerce data across multiple paginated webpages 证据：`examples/js/collect-paginated-ecommerce-data/README.md`
- **Example script: Collect news headlines across multiple paginated webpages**（documentation）：Example script: Collect news headlines across multiple paginated webpages 证据：`examples/js/collect-paginated-news-headlines/README.md`
- **Example script: collecting pricing data from e-commerce website using AgentQL**（documentation）：Example script: collecting pricing data from e-commerce website using AgentQL 证据：`examples/js/collect-pricing-data/README.md`
- **Example script: interacting with YouTube website and extracting video information using AgentQL**（documentation）：Example script: interacting with YouTube website and extracting video information using AgentQL 证据：`examples/js/collect-youtube-comments/README.md`
- **Example script: comparing product price across websites with AgentQL**（documentation）：Example script: comparing product price across websites with AgentQL 证据：`examples/js/compare-product-prices/README.md`
- **Example script: collecting pricing data from e-commerce website using AgentQL**（documentation）：Example script: collecting pricing data from e-commerce website using AgentQL 证据：`examples/js/first-steps/README.md`
- **Example script: interact with elements leveraging getByPrompt method**（documentation）：Example script: interact with elements leveraging getByPrompt method 证据：`examples/js/get-by-prompt/README.md`
- **Example script: use humanlike mouse movement to avoid antibot**（documentation）：Example script: use humanlike mouse movement to avoid antibot 证据：`examples/js/humanlike-antibot/README.md`
- **Example script: load additional content on page by scrolling**（documentation）：Example script: load additional content on page by scrolling 证据：`examples/js/infinite-scroll/README.md`
- **Example script: querying a list of items with AgentQL**（documentation）：Example script: querying a list of items with AgentQL 证据：`examples/js/list-query-usage/README.md`
- **Example script: login into websites**（documentation）：Example script: login into websites 证据：`examples/js/log-into-sites/README.md`
- **Example script: querying a list of items with AgentQL**（documentation）：Example script: querying a list of items with AgentQL 证据：`examples/js/maps_scraper/README.md`
- **Example script: perform sentiment analysis with AgentQL**（documentation）：Example script: perform sentiment analysis with AgentQL 证据：`examples/js/news-aggregator/README.md`
- **Example script: perform sentiment analysis with AgentQL**（documentation）：Example script: perform sentiment analysis with AgentQL 证据：`examples/js/perform-sentiment-analysis/README.md`
- **Example script: run the script in headless browser with AgentQL**（documentation）：Example script: run the script in headless browser with AgentQL 证据：`examples/js/run-script-in-headless-browser/README.md`
- **Example script: save and load authenticated state with AgentQL**（documentation）：Example script: save and load authenticated state with AgentQL 证据：`examples/js/save-and-load-authenticated-session/README.md`
- **Stealth mode: Running AgentQL in stealth mode and avoiding bot detection**（documentation）：Stealth mode: Running AgentQL in stealth mode and avoiding bot detection 证据：`examples/js/stealth-mode/README.md`
- **Example script: Automating form submission using AgentQL**（documentation）：Example script: Automating form submission using AgentQL 证据：`examples/js/submit-form/README.md`
- **Example script: Use existing browser with AgentQL**（documentation）：Example script: Use existing browser with AgentQL 证据：`examples/js/use-existing-browser/README.md`
- **Example script: Use remote browser with AgentQL**（documentation）：Example script: Use remote browser with AgentQL 证据：`examples/js/use-remote-browser/README.md`
- **Example script: wait for page to load with AgentQL**（documentation）：Example script: wait for page to load with AgentQL 证据：`examples/js/wait-for-entire-page-load/README.md`
- **Example script: get XPath of elements fetched with AgentQL**（documentation）：Example script: get XPath of elements fetched with AgentQL 证据：`examples/js/xpath/README.md`
- **Example script: close a cookie dialog with AgentQL**（documentation）：Example script: close a cookie dialog with AgentQL 证据：`examples/python/close_cookie_dialog/README.md`
- **Example script: closing popup windows with AgentQL**（documentation）：Example script: closing popup windows with AgentQL 证据：`examples/python/close_popup/README.md`
- **Example script: collecting pricing data with AgentQL from e-commerce websites**（documentation）：Example script: collecting pricing data with AgentQL from e-commerce websites 证据：`examples/python/collect_ecommerce_pricing_data/README.md`
- **Example script: Collect ecommerce data across multiple paginated webpages**（documentation）：Example script: Collect ecommerce data across multiple paginated webpages 证据：`examples/python/collect_paginated_ecommerce_listing_data/README.md`
- **Example script: Collect news headlines across multiple paginated webpages**（documentation）：Example script: Collect news headlines across multiple paginated webpages 证据：`examples/python/collect_paginated_news_headlines/README.md`
- **Example script: comparing product price across websites with AgentQL**（documentation）：Example script: comparing product price across websites with AgentQL 证据：`examples/python/compare_product_prices/README.md`
- **Example script: collecting pricing data from e-commerce website using AgentQL**（documentation）：Example script: collecting pricing data from e-commerce website using AgentQL 证据：`examples/python/first_steps/README.md`
- **Example script: interact with elements leveraging get by prompt method**（documentation）：Example script: interact with elements leveraging get by prompt method 证据：`examples/python/get_by_prompt/README.md`
- **Example script: use humanlike mouse movement to avoid antibot**（documentation）：Example script: use humanlike mouse movement to avoid antibot 证据：`examples/python/humanlike-antibot/README.md`
- **Example script: load additional content on page by scrolling**（documentation）：Example script: load additional content on page by scrolling 证据：`examples/python/infinite_scroll/README.md`
- **Example script: querying a list of items with AgentQL**（documentation）：Example script: querying a list of items with AgentQL 证据：`examples/python/list_query_usage/README.md`
- **Example script: logging into sites with AgentQL**（documentation）：Example script: logging into sites with AgentQL 证据：`examples/python/log_into_sites/README.md`
- **Example script: scraping Google Maps with AgentQL**（documentation）：Example script: scraping Google Maps with AgentQL 证据：`examples/python/maps_scraper/README.md`
- **Example script: News Aggregator to CSV**（documentation）：Example script: News Aggregator to CSV 证据：`examples/python/news-aggregator/README.md`
- **Example script: performing sentiment analysis with AgentQL**（documentation）：Example script: performing sentiment analysis with AgentQL 证据：`examples/python/perform_sentiment_analysis/README.md`
- **Example script: running a script in headless browser with AgentQL**（documentation）：Example script: running a script in headless browser with AgentQL 证据：`examples/python/run_script_in_headless_browser/README.md`
- **Example script: run AgentQL script in online development environment**（documentation）：Example script: run AgentQL script in online development environment 证据：`examples/python/run_script_online_in_google_colab/README.md`
- **Example script: saving and loading authenticated state with AgentQL**（documentation）：Example script: saving and loading authenticated state with AgentQL 证据：`examples/python/save_and_load_authenticated_session/README.md`
- **Stealth mode: Running AgentQL in stealth mode and avoiding bot detection**（documentation）：Stealth mode: Running AgentQL in stealth mode and avoiding bot detection 证据：`examples/python/stealth_mode/README.md`
- **Example script: Filling out and submitting a form using AgentQL**（documentation）：Example script: Filling out and submitting a form using AgentQL 证据：`examples/python/submit_form/README.md`
- **Example script: Use existing browser with AgentQL**（documentation）：Example script: Use existing browser with AgentQL 证据：`examples/python/use_existing_browser/README.md`
- **Example script: Use remote browser with AgentQL**（documentation）：Example script: Use remote browser with AgentQL 证据：`examples/python/use_remote_browser/README.md`
- **Example script: waiting for page to load with AgentQL**（documentation）：Example script: waiting for page to load with AgentQL 证据：`examples/python/wait_for_entire_page_load/README.md`
- **Example script: getting an element's XPath with AgentQL**（documentation）：Example script: getting an element's XPath with AgentQL 证据：`examples/python/xpath/README.md`
- **Package**（package_manifest）：{ "name": "javascript-sdk", "version": "1.0.0", "description": "AgentQL Examples for JavaScript SDK", "directories": { "example": "examples" }, "scripts": { "lint": "eslint --config .eslintrc.js .", "format": "prettier --write ." }, "author": "", "license": "MIT", "devDependencies": { "@trivago/prettier-plugin-sort-imports": "^4.3.0", "eslint": "^8.57.0", "eslint-config-prettier": "^9.1.0", "prettier": "^2.8.7" }, "dependencies": { "agentql": "latest", "openai": "^4.70.1", "playwright": "^1.48.2", "playwright-dompath": "^0.0.7" }, "overrides": { "axios": "^1.15.0", "flatted": "^3.4.2", "follow-redirects": "^1.16.0", "lodash": "^4.18.0", "minimatch": "^3.1.3" } } 证据：`examples/js/package.json`
- **License**（source_file）：Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files the “Software” , to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: 证据：`LICENSE`
- **Overview**（documentation）：Browse our opensource collection of useful example scripts and get started extracting data, scraping, and automating. 证据：`examples/examples.md`
- **Golden Images**（source_file）：golden images: nodejs: - tier: recommended alias: node-24-lts image: node:24-bookworm-slim uri: node:24-bookworm-slim@sha256:e8e2e91b1378f83c5b2dd15f0247f34110e2fe895f6ca7719dbb780f929368eb runtime version: "24.14.0" base os: Debian 12 Bookworm slim digest updated: "2026-02-27" eol: "2029-04-30" description: Node.js 24 LTS on Debian 12 Bookworm slim. DevOps-managed. Slim variant reduces attack surface vs the full image while retaining native-module compatibility unlike Alpine/musl . Receives timely Debian security patches for OS-level packages. use case: "All Node.js services, CI builds, tooling containers." - tier: acceptable alias: node-22-lts image: node:22-bookworm-slim uri: node:22-boo… 证据：`golden-images.yaml`
- **Template**（source_file）：async function main ⋮---- async function fetchData context, sessionUrl 证据：`.templates/js/template.js`
- **Main**（source_file）：{ "cells": { "cell type": "markdown", "metadata": {}, "source": "This example shows how to close a cookie dialog on a website using AgentQL with the Google Colaboratory https://colab.research.google.com/ environment." }, { "cell type": "markdown", "metadata": {}, "source": "Install AgentQL https://pypi.org/project/agentql/ library" }, { "cell type": "code", "execution count": null, "metadata": {}, "outputs": , "source": "%pip install agentql" }, { "cell type": "markdown", "metadata": {}, "source": "Install the Playwright dependency required by AgentQL." }, { "cell type": "code", "execution count": null, "metadata": {}, "outputs": , "source": "!playwright install chromium" }, { "cell type":… 证据：`examples/googlecolab/close_cookie_dialog/main.ipynb`
- **Main**（source_file）：{ "cells": { "cell type": "markdown", "metadata": {}, "source": "This example shows how to close popup windows like promotion form using AgentQL with the Google Colaboratory https://colab.research.google.com/ environment." }, { "cell type": "markdown", "metadata": {}, "source": "Install AgentQL https://pypi.org/project/agentql/ library" }, { "cell type": "code", "execution count": null, "metadata": {}, "outputs": , "source": "%pip install agentql" }, { "cell type": "markdown", "metadata": {}, "source": "Install the Playwright dependency required by AgentQL." }, { "cell type": "code", "execution count": null, "metadata": {}, "outputs": , "source": "!playwright install chromium" }, { "cell ty… 证据：`examples/googlecolab/close_popup/main.ipynb`
- **Main**（source_file）：{ "cells": { "cell type": "markdown", "metadata": {}, "source": "This example shows how to collect paginated data from an ecommerce website using AgentQL with the Google Colaboratory https://colab.research.google.com/ environment." }, { "cell type": "markdown", "metadata": {}, "source": "Install AgentQL https://pypi.org/project/agentql/ library" }, { "cell type": "code", "execution count": null, "metadata": {}, "outputs": , "source": "%pip install agentql" }, { "cell type": "markdown", "metadata": {}, "source": "Install the Playwright dependencency required by AgentQL." }, { "cell type": "code", "execution count": null, "metadata": {}, "outputs": , "source": "!playwright install chromium" }… 证据：`examples/googlecolab/collect_paginated_ecommerce_listing_data/main.ipynb`
- **Main**（source_file）：{ "cells": { "cell type": "markdown", "metadata": {}, "source": "This example shows how to collect paginated news headlines using AgentQL with the Google Colaboratory https://colab.research.google.com/ environment." }, { "cell type": "markdown", "metadata": {}, "source": "Install AgentQL https://pypi.org/project/agentql/ library" }, { "cell type": "code", "execution count": null, "metadata": {}, "outputs": , "source": "%pip install agentql" }, { "cell type": "markdown", "metadata": {}, "source": "Install the Playwright dependency required by AgentQL." }, { "cell type": "code", "execution count": null, "metadata": {}, "outputs": , "source": "!playwright install chromium" }, { "cell type": "m… 证据：`examples/googlecolab/collect_paginated_news_headlines/main.ipynb`
- **Main**（source_file）：{ "cells": { "cell type": "markdown", "metadata": {}, "source": "This example shows how to compare product prices using AgentQL with the Google Colaboratory https://colab.research.google.com/ environment." }, { "cell type": "markdown", "metadata": {}, "source": "Install AgentQL https://pypi.org/project/agentql/ library" }, { "cell type": "code", "execution count": null, "metadata": {}, "outputs": , "source": "%pip install agentql" }, { "cell type": "markdown", "metadata": {}, "source": "Install the Playwright dependency required by AgentQL." }, { "cell type": "code", "execution count": null, "metadata": {}, "outputs": , "source": "!playwright install chromium" }, { "cell type": "markdown",… 证据：`examples/googlecolab/compare_product_prices/main.ipynb`
- 其余 16 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`, `.templates/python/README.md`, `examples/js/close-cookie-dialog/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`, `.templates/python/README.md`, `examples/js/close-cookie-dialog/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。

- **AgentQL 项目介绍**：importance `high`
  - source_paths: README.md, examples/examples.md
- **快速入门指南**：importance `high`
  - source_paths: examples/python/first_steps/main.py, examples/js/first-steps/main.js, examples/python/run_script_online_in_google_colab/main.ipynb
- **Python SDK 详解**：importance `high`
  - source_paths: examples/python/first_steps/main.py, examples/python/collect_ecommerce_pricing_data/main.py, examples/python/submit_form/main.py
- **JavaScript SDK 详解**：importance `high`
  - source_paths: examples/js/package.json, examples/js/first-steps/main.js, examples/js/submit-form/main.js
- **浏览器调试器扩展**：importance `medium`
  - source_paths: examples/python/list_query_usage/README.md, examples/js/list-query-usage/README.md
- **Python 示例代码集**：importance `high`
  - source_paths: examples/python/collect_paginated_ecommerce_listing_data/main.py, examples/python/maps_scraper/main.py, examples/python/log_into_sites/main.py, examples/python/stealth_mode/main.py
- **JavaScript 示例代码集**：importance `high`
  - source_paths: examples/js/collect-paginated-ecommerce-data/main.js, examples/js/collect-youtube-comments/main.js, examples/js/maps_scraper/main.js
- **高级使用模式**：importance `medium`
  - source_paths: examples/python/use_remote_browser/main.py, examples/python/use_existing_browser/main.py, examples/python/stealth_mode/main.py, examples/python/save_and_load_authenticated_session/main.py, examples/js/use-remote-browser/main.js

## Repo Inspection Evidence / 源码检查证据

- repo_clone_verified: true
- repo_inspection_verified: true
- repo_commit: `cfc4193ea0f21d46fcea167789a3969edc6af5b9`
- inspected_files: `README.md`, `examples/examples.md`, `examples/js/package-lock.json`, `examples/js/.prettierrc.js`, `examples/js/package.json`, `examples/js/.eslintrc.js`, `examples/python/pyproject.toml`, `examples/python/poetry.toml`, `examples/js/list-query-usage/README.md`, `examples/js/list-query-usage/main.js`, `examples/js/log-into-sites/README.md`, `examples/js/log-into-sites/main.js`, `examples/js/close-cookie-dialog/README.md`, `examples/js/close-cookie-dialog/main.js`, `examples/js/stealth-mode/README.md`, `examples/js/stealth-mode/main.js`, `examples/js/infinite-scroll/README.md`, `examples/js/infinite-scroll/main.js`, `examples/js/run-script-in-headless-browser/README.md`, `examples/js/run-script-in-headless-browser/main.js`

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：Dependency Dashboard

- Trigger: GitHub 社区证据显示该项目存在一个安装相关的待验证问题：Dependency Dashboard
- 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_761b694cc0e94100b46ba5683041137b | https://github.com/tinyfish-io/agentql/issues/114 | 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 2: 来源证据：Starlog published a deep-dive on tinyfish-io/agentql

- Trigger: GitHub 社区证据显示该项目存在一个安装相关的待验证问题：Starlog published a deep-dive on tinyfish-io/agentql
- Host AI rule: 来源显示可能已有修复、规避或版本变化，说明书中必须标注适用版本。
- Why it matters: 可能增加新用户试用和生产接入成本。
- Evidence: community_evidence:github | cevd_55a8aa1466634fb39e0b679f753270ec | https://github.com/tinyfish-io/agentql/issues/148 | 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 | github_repo:760722197 | https://github.com/tinyfish-io/agentql | 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: 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 | github_repo:760722197 | https://github.com/tinyfish-io/agentql | 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 5: 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 | github_repo:760722197 | https://github.com/tinyfish-io/agentql | 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 6: 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 | github_repo:760722197 | https://github.com/tinyfish-io/agentql | 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: 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 | github_repo:760722197 | https://github.com/tinyfish-io/agentql | 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 8: 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 | github_repo:760722197 | https://github.com/tinyfish-io/agentql | 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。