# my-static-site - 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 my-static-site 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

- **正在使用 Claude/Codex/Cursor/Gemini 等宿主 AI 的开发者**：README 或插件配置提到多个宿主 AI。 证据：`README.md` Claim：`clm_0019` supported 0.86
- **希望把专业流程带进宿主 AI 的用户**：仓库包含 Skill 文档。 证据：`.agents/skills/complete-partial-pr/SKILL.md`, `.claude/skills/address-feedback/SKILL.md`, `.claude/skills/pre-push-review/SKILL.md`, `.claude/skills/testing-skill/SKILL.md` 等 Claim：`clm_0020` supported 0.86

## 它能做什么

- **Agent Framework**（可做安装前预览）：Core framework for building AI agents with Pydantic validation, type safety, and model-agnostic design. 证据：`README.md`, `pydantic_ai_slim/pydantic_ai/.agents/skills/building-pydantic-ai-agents/SKILL.md`, `AGENTS.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Multi-Model Provider Support**（需要安装后验证）：Model-agnostic support for dozens of LLM providers including OpenAI, Anthropic, Gemini, DeepSeek, Grok, Cohere, Mistral, Azure, Bedrock, and local models. 证据：`README.md`, `docs/install.md`, `clai/README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0004` supported 0.86, `clm_0006` supported 0.86 等
- **Tool Calling**（可做安装前预览）：Function calling capability allowing agents to invoke Python functions with structured arguments and dependency injection. 证据：`pydantic_ai_slim/pydantic_ai/.agents/skills/building-pydantic-ai-agents/SKILL.md`, `docs/tools.md`, `pydantic_ai_slim/pydantic_ai/toolsets/AGENTS.md` Claim：`clm_0001` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86, `clm_0005` supported 0.86 等
- **Structured Output**（可做安装前预览）：Output validation using Pydantic models, enabling type-safe response parsing and schema enforcement. 证据：`pydantic_ai_slim/pydantic_ai/.agents/skills/building-pydantic-ai-agents/SKILL.md`, `README.md`, `docs/output.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Dependency Injection**（可做安装前预览）：Context management system for injecting dependencies (database connections, user context, configuration) into agent tools and instructions. 证据：`pydantic_ai_slim/pydantic_ai/.agents/skills/building-pydantic-ai-agents/SKILL.md`, `docs/dependencies.md`, `pydantic_ai_slim/pydantic_ai/AGENTS.md` Claim：`clm_0001` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86, `clm_0005` supported 0.86 等
- **Streaming**（可做安装前预览）：Real-time streaming of agent events, text, and tool calls for responsive user interfaces. 证据：`README.md`, `clai/README.md`, `docs/streaming.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0004` supported 0.86, `clm_0006` supported 0.86 等
- **Capabilities System**（可做安装前预览）：Composable units of agent behavior (thinking, web search, MCP) that bundle tools, hooks, instructions, and settings. 证据：`pydantic_ai_slim/pydantic_ai/.agents/skills/building-pydantic-ai-agents/SKILL.md`, `README.md`, `pydantic_ai_slim/pydantic_ai/capabilities/AGENTS.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Model Context Protocol (MCP)**（可做安装前预览）：Integration with the Model Context Protocol for connecting to MCP servers and tools. 证据：`README.md`, `docs/mcp/overview.md`, `docs/install.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0004` supported 0.86, `clm_0006` supported 0.86 等
- **Agent2Agent (A2A) Protocol**（可做安装前预览）：Protocol support for inter-agent communication and delegation. 证据：`README.md`, `docs/a2a.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0004` supported 0.86, `clm_0006` supported 0.86 等
- **Observability with Pydantic Logfire**（可做安装前预览）：OpenTelemetry-based observability integration for tracing, debugging, and performance monitoring. 证据：`README.md`, `docs/install.md`, `pydantic_evals/README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0004` supported 0.86, `clm_0006` supported 0.86 等
- **CLI (clai)**（需要安装后验证）：Command-line interface for interactive LLM chat, web-based chat UI, and agent execution. 证据：`clai/README.md`, `clai/README.md`, `clai/README.md` Claim：`clm_0002` supported 0.86, `clm_0006` supported 0.86, `clm_0011` supported 0.86, `clm_0012` supported 0.86 等
- **Web Chat UI**（可做安装前预览）：Browser-based chat interface for interacting with agents, served via clai web command. 证据：`clai/README.md`, `clai/README.md`, `pydantic_ai_slim/pydantic_ai/ui/AGENTS.md` Claim：`clm_0002` supported 0.86, `clm_0006` supported 0.86, `clm_0011` supported 0.86, `clm_0012` supported 0.86 等
- **Pydantic Graph (State Machines)**（可做安装前预览）：Graph-based state machine library for defining workflows using Python syntax with typed edges. 证据：`pydantic_graph/README.md`, `pydantic_graph/README.md`, `docs/graph.md` Claim：`clm_0013` supported 0.86
- **Pydantic Evals**（可做安装前预览）：Evaluation framework for testing and monitoring AI agent performance with OpenTelemetry tracing. 证据：`pydantic_evals/README.md`, `pydantic_evals/README.md`, `README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0004` supported 0.86, `clm_0006` supported 0.86 等
- **Durable Execution**（需要安装后验证）：Integration with durable execution engines (Temporal, DBOS, Prefect, Restate) for resilient agent runs. 证据：`pydantic_ai_slim/pydantic_ai/durable_exec/AGENTS.md`, `pydantic_ai_slim/pydantic_ai/durable_exec/AGENTS.md` Claim：`clm_0015` supported 0.86
- **Agent Spec (YAML/JSON)**（可做安装前预览）：Define agents entirely in YAML or JSON specification files without writing Python code. 证据：`README.md`, `docs/agent-spec.md`, `clai/README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0004` supported 0.86, `clm_0006` supported 0.86 等
- **Native Tools**（可做安装前预览）：Provider-native tool support (web search, web fetch, image generation) with local fallback capabilities. 证据：`pydantic_ai_slim/pydantic_ai/native_tools/AGENTS.md`, `README.md`, `docs/native-tools.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0004` supported 0.86, `clm_0006` supported 0.86 等
- **Embeddings/Vector Search**（可做安装前预览）：Support for embedding models including sentence transformers and VoyageAI for vector-based search. 证据：`docs/install.md`, `docs/install.md`, `docs/embeddings.md` Claim：`clm_0002` supported 0.86, `clm_0008` supported 0.86, `clm_0010` supported 0.86, `clm_0018` supported 0.86

## 怎么开始

- `pip install 'pydantic-evals[logfire]'` 证据：`docs/evals/evaluators/span-based.md` Claim：`clm_0021` unverified 0.25
- `pip install pydantic-evals` 证据：`docs/evals/quick-start.md` Claim：`clm_0021` unverified 0.25, `clm_0022` unverified 0.25
- `git clone https://github.com/ag-ui-protocol/ag-ui.git` 证据：`docs/examples/ag-ui.md` Claim：`clm_0023` unverified 0.25
- `curl -X POST <webhook endpoint URL> \` 证据：`docs/examples/slack-lead-qualifier.md` Claim：`clm_0024` unverified 0.25
- `pip install gradio>=6.7.0` 证据：`docs/examples/weather-agent.md` Claim：`clm_0025` unverified 0.25
- `pip install pydantic-graph` 证据：`docs/graph/builder/index.md` Claim：`clm_0026` unverified 0.25
- `pip install pydantic-ai` 证据：`docs/graph/builder/index.md` Claim：`clm_0027` unverified 0.25
- `uv tool install clai` 证据：`docs/cli.md` Claim：`clm_0028` supported 0.86
- `pip install clai` 证据：`docs/cli.md` Claim：`clm_0029` supported 0.86
- `npx skills add pydantic/skills` 证据：`docs/coding-agent-skills.md` Claim：`clm_0030` supported 0.86

## 继续前判断卡

- **当前建议**：先做权限沙盒试用
- **为什么**：项目存在安装命令、宿主配置或本地写入线索，不建议直接进入主力环境，应先在隔离环境试装。

### 30 秒判断

- **现在怎么做**：先做权限沙盒试用
- **最小安全下一步**：先跑 Prompt Preview；若仍要安装，只在隔离环境试装
- **先别相信**：工具权限边界不能在安装前相信。
- **继续会触碰**：命令执行、宿主 AI 配置、本地环境或项目文件

### 现在可以相信

- **适合人群线索：正在使用 Claude/Codex/Cursor/Gemini 等宿主 AI 的开发者**（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
- **适合人群线索：希望把专业流程带进宿主 AI 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：`.agents/skills/complete-partial-pr/SKILL.md`, `.claude/skills/address-feedback/SKILL.md`, `.claude/skills/pre-push-review/SKILL.md`, `.claude/skills/testing-skill/SKILL.md` 等 Claim：`clm_0020` supported 0.86
- **能力存在：Agent Framework**（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`, `pydantic_ai_slim/pydantic_ai/.agents/skills/building-pydantic-ai-agents/SKILL.md`, `AGENTS.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：Multi-Model Provider Support**（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`, `docs/install.md`, `clai/README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0004` supported 0.86, `clm_0006` supported 0.86
- **能力存在：Tool Calling**（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：`pydantic_ai_slim/pydantic_ai/.agents/skills/building-pydantic-ai-agents/SKILL.md`, `docs/tools.md`, `pydantic_ai_slim/pydantic_ai/toolsets/AGENTS.md` Claim：`clm_0001` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86, `clm_0005` supported 0.86
- **能力存在：Structured Output**（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：`pydantic_ai_slim/pydantic_ai/.agents/skills/building-pydantic-ai-agents/SKILL.md`, `README.md`, `docs/output.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86

### 现在还不能相信

- **工具权限边界不能在安装前相信。**（unverified）：MCP/tool 类项目通常会触碰文件、网络、浏览器或外部 API，必须真实检查权限和日志。
- **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。 证据：`.agents/skills/complete-partial-pr/SKILL.md`, `.claude/skills/address-feedback/SKILL.md`, `.claude/skills/pre-push-review/SKILL.md`, `.claude/skills/testing-skill/SKILL.md` 等
- **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。
- **安装命令是否需要网络、权限或全局写入？**（unverified）：这影响企业环境和个人环境的安装风险。 证据：`docs/evals/evaluators/span-based.md`

### 继续会触碰什么

- **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：`docs/cli.md`, `docs/coding-agent-skills.md`, `docs/contributing.md`, `docs/evals/evaluators/span-based.md` 等
- **宿主 AI 配置**：Claude/Codex/Cursor/Gemini/OpenCode 等宿主的 plugin、Skill 或规则加载配置。 原因：宿主配置会改变 AI 后续工作方式，可能和用户已有规则冲突。 证据：`.agents/skills/complete-partial-pr/SKILL.md`, `.claude/skills/address-feedback/SKILL.md`, `.claude/skills/pre-push-review/SKILL.md`, `.claude/skills/testing-skill/SKILL.md` 等
- **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`, `clai/README.md`, `docs/install.md`, `pydantic_ai_slim/pydantic_ai/durable_exec/AGENTS.md`
- **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**：用安装前交互式试用判断工作方式是否匹配，不需要授权或改环境。（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。）
- **先备份宿主 AI 配置**：Skill、plugin、规则文件可能改变 Claude/Cursor/Codex 的默认行为。（适用：存在插件 manifest、Skill 或宿主规则入口时。）
- **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。
- **准备移除宿主 plugin / Skill / 规则入口**：如果试装后行为异常，可以把宿主 AI 恢复到试装前状态。
- **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_0035` 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：`docs/cli.md`, `docs/coding-agent-skills.md`, `docs/contributing.md`, `docs/evals/evaluators/span-based.md` 等 Claim：`clm_0036` supported 0.86
- **风险**： 处理方式：
- **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

- **Agent Framework**：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`, `pydantic_ai_slim/pydantic_ai/.agents/skills/building-pydantic-ai-agents/SKILL.md`, `AGENTS.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Multi-Model Provider 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：`README.md`, `docs/install.md`, `clai/README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0004` supported 0.86, `clm_0006` supported 0.86 等
- **Tool Calling**：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：`pydantic_ai_slim/pydantic_ai/.agents/skills/building-pydantic-ai-agents/SKILL.md`, `docs/tools.md`, `pydantic_ai_slim/pydantic_ai/toolsets/AGENTS.md` Claim：`clm_0001` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86, `clm_0005` supported 0.86 等
- **Structured Output**：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：`pydantic_ai_slim/pydantic_ai/.agents/skills/building-pydantic-ai-agents/SKILL.md`, `README.md`, `docs/output.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Dependency Injection**：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：`pydantic_ai_slim/pydantic_ai/.agents/skills/building-pydantic-ai-agents/SKILL.md`, `docs/dependencies.md`, `pydantic_ai_slim/pydantic_ai/AGENTS.md` Claim：`clm_0001` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86, `clm_0005` supported 0.86 等
- **Streaming**：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`, `clai/README.md`, `docs/streaming.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0004` supported 0.86, `clm_0006` supported 0.86 等
- **Capabilities System**：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：`pydantic_ai_slim/pydantic_ai/.agents/skills/building-pydantic-ai-agents/SKILL.md`, `README.md`, `pydantic_ai_slim/pydantic_ai/capabilities/AGENTS.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Model Context Protocol (MCP)**：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`, `docs/mcp/overview.md`, `docs/install.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0004` supported 0.86, `clm_0006` supported 0.86 等
- **Agent2Agent (A2A) Protocol**：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`, `docs/a2a.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0004` supported 0.86, `clm_0006` supported 0.86 等
- **Observability with Pydantic Logfire**：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`, `docs/install.md`, `pydantic_evals/README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0004` supported 0.86, `clm_0006` supported 0.86 等
- **CLI (clai)**：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：`clai/README.md`, `clai/README.md`, `clai/README.md` Claim：`clm_0002` supported 0.86, `clm_0006` supported 0.86, `clm_0011` supported 0.86, `clm_0012` supported 0.86 等
- **Web Chat UI**：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：`clai/README.md`, `clai/README.md`, `pydantic_ai_slim/pydantic_ai/ui/AGENTS.md` Claim：`clm_0002` supported 0.86, `clm_0006` supported 0.86, `clm_0011` supported 0.86, `clm_0012` supported 0.86 等
- **Pydantic Graph (State Machines)**：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：`pydantic_graph/README.md`, `pydantic_graph/README.md`, `docs/graph.md` Claim：`clm_0013` supported 0.86
- **Pydantic Evals**：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：`pydantic_evals/README.md`, `pydantic_evals/README.md`, `README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0004` supported 0.86, `clm_0006` supported 0.86 等
- **Durable 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：`pydantic_ai_slim/pydantic_ai/durable_exec/AGENTS.md`, `pydantic_ai_slim/pydantic_ai/durable_exec/AGENTS.md` Claim：`clm_0015` supported 0.86
- **Agent Spec (YAML/JSON)**：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`, `docs/agent-spec.md`, `clai/README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0004` supported 0.86, `clm_0006` supported 0.86 等
- **Native Tools**：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：`pydantic_ai_slim/pydantic_ai/native_tools/AGENTS.md`, `README.md`, `docs/native-tools.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0004` supported 0.86, `clm_0006` supported 0.86 等
- **Embeddings/Vector Search**：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：`docs/install.md`, `docs/install.md`, `docs/embeddings.md` Claim：`clm_0002` supported 0.86, `clm_0008` supported 0.86, `clm_0010` supported 0.86, `clm_0018` supported 0.86

### 上下文规模

- 文件总数：612
- 重要文件覆盖：40/612
- 证据索引条目：80
- 角色 / Skill 条目：5

### 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 my-static-site 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 my-static-site 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 my-static-site 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 5 个角色 / Skill / 项目文档条目。

- **complete-partial-pr**（skill）：Evaluate and complete an issue or PR where the submitted patch fixes only a narrow symptom of the reported pain point. Use when a contribution may miss adjacent integration surfaces, provider/spec semantics, roundtrip behavior, tests, docs, or historical maintainer decisions. 激活提示：当用户任务与“complete-partial-pr”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`.agents/skills/complete-partial-pr/SKILL.md`
- **address-feedback**（skill）：Find and address unresolved PR review comments for the current branch, then summarize the changes and help reply to and resolve threads after user approval. 激活提示：当用户任务与“address-feedback”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`.claude/skills/address-feedback/SKILL.md`
- **Pre-push Review**（skill）：Review the current branch against main, simulating the automated CI review from the bots workflow 激活提示：当用户任务与“Pre-push Review”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`.claude/skills/pre-push-review/SKILL.md`
- **testing-skill**（skill）：Record, rewrite, and debug VCR cassettes for HTTP recordings. Use when running tests with --record-mode, verifying cassette playback, or inspecting request/response bodies in YAML cassettes. 激活提示：当用户任务与“testing-skill”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`.claude/skills/testing-skill/SKILL.md`
- **building-pydantic-ai-agents**（skill）：Build AI agents with Pydantic AI — tools, capabilities, structured output, streaming, testing, and multi-agent patterns. Use when the user mentions Pydantic AI, imports pydantic ai, or asks to build an AI agent, add tools/capabilities, stream output, define agents from YAML, or test agent behavior. 激活提示：当用户任务与“building-pydantic-ai-agents”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`pydantic_ai_slim/pydantic_ai/.agents/skills/building-pydantic-ai-agents/SKILL.md`

## 证据索引

- 共索引 80 条证据。

- **docs/ Guidelines**（documentation）：- Link all concepts, features, and API elements to their docs/reference pages using anchor fragments section-name for specific sections — Improves discoverability and reduces user friction by providing direct navigation to relevant documentation context - Use reference-style links for API elements: ElementName module.path.ElementName — enables hover docs and navigation in mkdocs — Provides interactive documentation features like tooltips and jump-to-definition that plain backticks cannot support - Omit deprecated features from user-facing docs — document only current approaches — Prevents users from learning outdated patterns and reduces confusion about the recommended way forward - Write p… 证据：`docs/AGENTS.md`
- **How we work — the short version**（documentation）：We'd love you to contribute to Pydantic AI! 证据：`docs/contributing.md`
- **Installation**（documentation）：Pydantic AI is available on PyPI as pydantic-ai https://pypi.org/project/pydantic-ai/ so installation is as simple as: 证据：`docs/install.md`
- **Your primary responsibility is to the project and its users**（documentation）：Welcome to the repository for Pydantic AI https://ai.pydantic.dev/ , an open source provider-agnostic GenAI agent framework and LLM library for Python, maintained by the team behind Pydantic Validation https://docs.pydantic.dev/ and Pydantic Logfire https://docs.pydantic.dev/logfire/ . 证据：`AGENTS.md`
- **Pydantic AI is a Python agent framework designed to help you quickly, confidently, and painlessly build production grad…**（documentation）：GenAI Agent Framework, the Pydantic way 证据：`README.md`
- **clai**（documentation）：! CI https://github.com/pydantic/pydantic-ai/actions/workflows/ci.yml/badge.svg?event=push https://github.com/pydantic/pydantic-ai/actions/workflows/ci.yml?query=branch%3Amain ! Coverage https://coverage-badge.samuelcolvin.workers.dev/pydantic/pydantic-ai.svg https://coverage-badge.samuelcolvin.workers.dev/redirect/pydantic/pydantic-ai ! PyPI https://img.shields.io/pypi/v/clai.svg https://pypi.python.org/pypi/clai ! versions https://img.shields.io/pypi/pyversions/clai.svg https://github.com/pydantic/pydantic-ai ! license https://img.shields.io/github/license/pydantic/pydantic-ai.svg?v https://github.com/pydantic/pydantic-ai/blob/main/LICENSE 证据：`clai/README.md`
- **docs site**（documentation）：This is a cloudflare workers static assets https://developers.cloudflare.com/workers/static-assets/ site used to host the documentation. 证据：`docs-site/README.md`
- **Pydantic AI Examples**（documentation）：! CI https://github.com/pydantic/pydantic-ai/actions/workflows/ci.yml/badge.svg?event=push https://github.com/pydantic/pydantic-ai/actions/workflows/ci.yml?query=branch%3Amain ! Coverage https://coverage-badge.samuelcolvin.workers.dev/pydantic/pydantic-ai.svg https://coverage-badge.samuelcolvin.workers.dev/redirect/pydantic/pydantic-ai ! PyPI https://img.shields.io/pypi/v/pydantic-ai.svg https://pypi.python.org/pypi/pydantic-ai ! versions https://img.shields.io/pypi/pyversions/pydantic-ai.svg https://github.com/pydantic/pydantic-ai ! license https://img.shields.io/github/license/pydantic/pydantic-ai.svg?v https://github.com/pydantic/pydantic-ai/blob/main/LICENSE 证据：`examples/README.md`
- **Pydantic AI Slim**（documentation）：! CI https://github.com/pydantic/pydantic-ai/actions/workflows/ci.yml/badge.svg?event=push https://github.com/pydantic/pydantic-ai/actions/workflows/ci.yml?query=branch%3Amain ! Coverage https://coverage-badge.samuelcolvin.workers.dev/pydantic/pydantic-ai.svg https://coverage-badge.samuelcolvin.workers.dev/redirect/pydantic/pydantic-ai ! PyPI https://img.shields.io/pypi/v/pydantic-ai-slim.svg https://pypi.python.org/pypi/pydantic-ai-slim ! versions https://img.shields.io/pypi/pyversions/pydantic-ai-slim.svg https://github.com/pydantic/pydantic-ai ! license https://img.shields.io/github/license/pydantic/pydantic-ai.svg?v https://github.com/pydantic/pydantic-ai/blob/main/LICENSE 证据：`pydantic_ai_slim/README.md`
- **Pydantic Evals**（documentation）：! CI https://github.com/pydantic/pydantic-ai/actions/workflows/ci.yml/badge.svg?event=push https://github.com/pydantic/pydantic-ai/actions/workflows/ci.yml?query=branch%3Amain ! Coverage https://coverage-badge.samuelcolvin.workers.dev/pydantic/pydantic-ai.svg https://coverage-badge.samuelcolvin.workers.dev/redirect/pydantic/pydantic-ai ! PyPI https://img.shields.io/pypi/v/pydantic-evals.svg https://pypi.python.org/pypi/pydantic-evals ! python versions https://img.shields.io/pypi/pyversions/pydantic-evals.svg https://github.com/pydantic/pydantic-ai ! license https://img.shields.io/github/license/pydantic/pydantic-ai.svg https://github.com/pydantic/pydantic-ai/blob/main/LICENSE 证据：`pydantic_evals/README.md`
- **Pydantic Graph**（documentation）：! CI https://github.com/pydantic/pydantic-ai/actions/workflows/ci.yml/badge.svg?event=push https://github.com/pydantic/pydantic-ai/actions/workflows/ci.yml?query=branch%3Amain ! Coverage https://coverage-badge.samuelcolvin.workers.dev/pydantic/pydantic-ai.svg https://coverage-badge.samuelcolvin.workers.dev/redirect/pydantic/pydantic-ai ! PyPI https://img.shields.io/pypi/v/pydantic-graph.svg https://pypi.python.org/pypi/pydantic-graph ! python versions https://img.shields.io/pypi/pyversions/pydantic-graph.svg https://github.com/pydantic/pydantic-ai ! license https://img.shields.io/github/license/pydantic/pydantic-ai.svg https://github.com/pydantic/pydantic-ai/blob/main/LICENSE 证据：`pydantic_graph/README.md`
- **pydantic ai slim/pydantic ai/ Guidelines**（documentation）：pydantic ai slim/pydantic ai/ Guidelines 证据：`pydantic_ai_slim/pydantic_ai/AGENTS.md`
- **capabilities/ Guidelines**（documentation）：Capabilities are the composable home for cross-cutting agent behavior. 证据：`pydantic_ai_slim/pydantic_ai/capabilities/AGENTS.md`
- **durable exec/ Guidelines**（documentation）：Durable execution integrations are first-class compatibility targets. 证据：`pydantic_ai_slim/pydantic_ai/durable_exec/AGENTS.md`
- **pydantic ai slim/pydantic ai/models/ Guidelines**（documentation）：pydantic ai slim/pydantic ai/models/ Guidelines 证据：`pydantic_ai_slim/pydantic_ai/models/AGENTS.md`
- **native tools/ Guidelines**（documentation）：- Every native tool must have a corresponding capability extending NativeOrLocalTool in capabilities/ — capabilities are the primary user-facing API for enabling tool features on agents; a native tool without a capability is undiscoverable for users working with the capabilities list - Local fallback e.g., WebSearch , WebFetch : capability falls back to a function tool on providers without native support - Subagent fallback e.g., ImageGeneration , XSearch : capability delegates to a subagent running another provider's model via fallback model - When a provider's API has request-level parameters controlling raw tool output inclusion e.g., xAI include , OpenAI include , expose the tool-specif… 证据：`pydantic_ai_slim/pydantic_ai/native_tools/AGENTS.md`
- **profiles/ Guidelines**（documentation）：Profiles describe model-family capability facts and schema/request quirks. 证据：`pydantic_ai_slim/pydantic_ai/profiles/AGENTS.md`
- **providers/ Guidelines**（documentation）：Providers own API clients, authentication, base URLs, HTTP lifecycle, and provider-level model/profile inference. 证据：`pydantic_ai_slim/pydantic_ai/providers/AGENTS.md`
- **toolsets/ Guidelines**（documentation）：Toolsets are reusable tool collections with lifecycle, instructions, and execution boundaries. 证据：`pydantic_ai_slim/pydantic_ai/toolsets/AGENTS.md`
- **Backwards compatibility in UI adapters specially AG-UI**（documentation）：Backwards compatibility in UI adapters specially AG-UI 证据：`pydantic_ai_slim/pydantic_ai/ui/AGENTS.md`
- **Package**（package_manifest）：{ "name": "my-static-site", "version": "0.0.0", "private": true, "scripts": { "typecheck": "tsgo --noEmit", "typegen": "wrangler types --strict-vars false --include-runtime false", "deploy": "wrangler deploy", "dev": "wrangler dev" }, "dependencies": { "@pydantic/logfire-api": "^0.8.2", "@pydantic/logfire-cf-workers": "^0.8.2", "marked": "^15.0.8" }, "devDependencies": { "@cloudflare/workers-types": "^4.20251004.0", "@typescript/native-preview": "^7.0.0-dev.20251006.1", "wrangler": "^4.59.1" }, "overrides": { "protobufjs": "^7.5.8", "@protobufjs/utf8": "^1.1.1" } } 证据：`docs-site/package.json`
- **Complete Partial PR**（skill_instruction）：Use this when a PR or issue patch fixes a small visible failure but may not address the full integration contract. The goal is to turn a narrow contribution into a maintainable Pydantic AI change, or to explain precisely why it should stay narrow. 证据：`.agents/skills/complete-partial-pr/SKILL.md`
- **Address PR Review Feedback**（skill_instruction）：Find and address all review comments on the PR for the current branch. For each comment: 证据：`.claude/skills/address-feedback/SKILL.md`
- **Pre-push Review**（skill_instruction）：Simulate the automated CI review job locally before pushing or opening a PR. 证据：`.claude/skills/pre-push-review/SKILL.md`
- **Building AI Agents with Pydantic AI**（skill_instruction）：Building AI Agents with Pydantic AI 证据：`pydantic_ai_slim/pydantic_ai/.agents/skills/building-pydantic-ai-agents/SKILL.md`
- **Pytest VCR Workflow**（skill_instruction）：Use this skill when recording or re-recording VCR cassettes for tests, or when debugging cassette contents. 证据：`.claude/skills/testing-skill/SKILL.md`
- **License**（source_file）：Copyright c Pydantic Services Inc. 2024 to present 证据：`LICENSE`
- **License**（source_file）：Copyright c Pydantic Services Inc. 2024 to present 证据：`clai/LICENSE`
- **License**（source_file）：Copyright c Pydantic Services Inc. 2024 to present 证据：`examples/LICENSE`
- **License**（source_file）：Copyright c Pydantic Services Inc. 2024 to present 证据：`pydantic_ai_slim/LICENSE`
- **License**（source_file）：Copyright c Pydantic Services Inc. 2024 to present 证据：`pydantic_evals/LICENSE`
- **License**（source_file）：Copyright c Pydantic Services Inc. 2024 to present 证据：`pydantic_graph/LICENSE`
- **Agent2Agent A2A Protocol**（documentation）：!!! warning "Deprecated in 1.x, removed in 2.0" Agent.to a2a and the pydantic-ai-slim a2a extra are deprecated and will be removed in 2.0. The fasta2a package is now maintained at datalayer/fasta2a https://github.com/datalayer/fasta2a and ships a Pydantic AI bridge since v0.6.1 https://github.com/datalayer/fasta2a/releases/tag/v0.6.1 . Install it with the pydantic-ai extra and use agent to a2a directly: 证据：`docs/a2a.md`
- **Agent Specs**（documentation）：Agent specs let you define agents declaratively in YAML or JSON — model models/overview.md , instructions agent.md instructions , capabilities capabilities.md , and all. One line to load, no Python agent construction code required. 证据：`docs/agent-spec.md`
- **Introduction**（documentation）：Agents are Pydantic AI's primary interface for interacting with LLMs. 证据：`docs/agent.md`
- **Capabilities**（documentation）：A capability is a reusable, composable unit of agent behavior. Instead of threading multiple arguments through your Agent constructor — instructions agent.md instructions here, model settings agent.md model-run-settings there, a toolset toolsets.md somewhere else, a history processor message-history.md processing-message-history on yet another parameter — you can bundle related behavior into a single capability and pass it via the capabilities pydantic ai.agent.Agent. init parameter. 证据：`docs/capabilities.md`
- **Upgrade Guide**（documentation）：In September 2025, Pydantic AI reached V1, which means we're committed to API stability: we will not introduce changes that break your code until V2. For more information, review our Version Policy version-policy.md . 证据：`docs/changelog.md`
- **Command Line Interface CLI**（documentation）：Pydantic AI comes with a CLI, clai pronounced "clay" . You can use it to chat with various LLMs and quickly get answers, right from the command line, or spin up a uvicorn server to chat with your Pydantic AI agents from your browser. 证据：`docs/cli.md`
- **Coding Agent Skills**（documentation）：If you're building Pydantic AI applications with a coding agent, you can install the Pydantic AI skill from the pydantic/skills https://github.com/pydantic/skills repository to give your agent up-to-date framework knowledge. 证据：`docs/coding-agent-skills.md`
- **Common Tools**（documentation）：Pydantic AI ships with native tools that can be used to enhance your agent's capabilities. 证据：`docs/common-tools.md`
- **Deferred Tools**（documentation）：There are a few scenarios where the model should be able to call a tool that should not or cannot be executed during the same agent run inside the same Python process: 证据：`docs/deferred-tools.md`
- **Dependencies**（documentation）：Pydantic AI uses a dependency injection system to provide data and services to your agent's system prompts agent.md system-prompts , tools tools.md and output validators output.md output-validator-functions . 证据：`docs/dependencies.md`
- **Direct Model Requests**（documentation）：The direct module provides low-level methods for making imperative requests to LLMs where the only abstraction is input and output schema translation, enabling you to use all models with the same API. 证据：`docs/direct.md`
- **Embeddings**（documentation）：Embeddings are vector representations of text that capture semantic meaning. They're essential for building: 证据：`docs/embeddings.md`
- **Pydantic Evals**（documentation）：Pydantic Evals is a powerful evaluation framework for systematically testing and evaluating AI systems, from simple LLM calls to complex multi-agent applications. 证据：`docs/evals.md`
- **Extensibility**（documentation）：Pydantic AI is designed to be extended. Capabilities capabilities.md are the primary extension point — they bundle tools, lifecycle hooks, instructions, and model settings into reusable units that can be shared across agents, packaged as libraries, and loaded from spec files agent-spec.md . 证据：`docs/extensibility.md`
- **Pydantic AI Gateway**（documentation）：Pydantic AI Gateway https://logfire.pydantic.dev/ is a unified interface for accessing multiple AI providers with a single key, managed through Pydantic Logfire https://logfire.pydantic.dev/ . Features include built-in OpenTelemetry observability, real-time cost monitoring, failover management, and native integration with the other tools in the Pydantic stack https://pydantic.dev/ . 证据：`docs/gateway.md`
- **Graphs**（documentation）：!!! danger "Don't use a nail gun unless you need a nail gun" If Pydantic AI agents agent.md are a hammer, and multi-agent workflows multi-agent-applications.md are a sledgehammer, then graphs are a nail gun: 证据：`docs/graph.md`
- **Getting Help**（documentation）：If you need help getting started with Pydantic AI or with advanced usage, the following sources may be useful. 证据：`docs/help.md`
- **Hooks**（documentation）：Hooks let you intercept and modify agent behavior at every stage of a run — model requests, tool calls, streaming events — using simple decorators or constructor arguments. No subclassing needed. 证据：`docs/hooks.md`
- **Pydantic AI {.hide}**（documentation）：--8<-- "docs/.partials/index-header.html" 证据：`docs/index.md`
- **Image, Audio, Video & Document Input**（documentation）：Image, Audio, Video & Document Input 证据：`docs/input.md`
- **Pydantic Logfire Debugging and Monitoring**（documentation）：Pydantic Logfire Debugging and Monitoring 证据：`docs/logfire.md`
- **Messages and chat history**（documentation）：Pydantic AI provides access to messages exchanged during an agent run. These messages can be used both to continue a coherent conversation, and to understand how an agent performed. 证据：`docs/message-history.md`
- **Multi-agent Applications**（documentation）：There are roughly five levels of complexity when building applications with Pydantic AI: 证据：`docs/multi-agent-applications.md`
- **Native Tools**（documentation）：Native tools are native tools provided by LLM providers that can be used to enhance your agent's capabilities. Unlike common tools common-tools.md , which are custom implementations that Pydantic AI executes, native tools are executed directly by the model provider. 证据：`docs/native-tools.md`
- **city='London' country='United Kingdom'**（documentation）："Output" refers to the final value returned from running an agent agent.md running-agents . This can be either plain text, structured data structured-output , an image image-output , or the result of a function output-functions called with arguments provided by the model. 证据：`docs/output.md`
- **HTTP Request Retries**（documentation）：Pydantic AI provides retry functionality for HTTP requests made by model providers through custom HTTP transports. This is particularly useful for handling transient failures like rate limits, network timeouts, or temporary server errors. 证据：`docs/retries.md`
- **Thinking**（documentation）：Thinking or reasoning is the process by which a model works through a problem step-by-step before providing its final answer. 证据：`docs/thinking.md`
- **Third-Party Tools**（documentation）：Pydantic AI supports integration with various third-party tool libraries, allowing you to leverage existing tool ecosystems in your agents. Third-party tools are also available as capabilities capabilities.md third-party-capabilities — see Extensibility extensibility.md for the full ecosystem. 证据：`docs/third-party-tools.md`
- 其余 20 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：`docs/AGENTS.md`, `docs/contributing.md`, `docs/install.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：`docs/AGENTS.md`, `docs/contributing.md`, `docs/install.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。

- **PydanticAI 概述**：importance `high`
  - source_paths: README.md, docs/index.md, pydantic_ai_slim/pydantic_ai/__init__.py
- **安装与快速入门**：importance `high`
  - source_paths: docs/install.md, pydantic_ai_slim/pyproject.toml
- **系统架构**：importance `high`
  - source_paths: pydantic_ai_slim/pydantic_ai/run.py, pydantic_ai_slim/pydantic_ai/agent/__init__.py, pydantic_ai_slim/pydantic_ai/agent/abstract.py, pydantic_ai_slim/pydantic_ai/messages.py, pydantic_ai_slim/pydantic_ai/result.py
- **代理系统详解**：importance `high`
  - source_paths: pydantic_ai_slim/pydantic_ai/agent/__init__.py, pydantic_ai_slim/pydantic_ai/agent/spec.py, pydantic_ai_slim/pydantic_ai/agent/wrapper.py, docs/agent.md, docs/api/agent.md
- **工具系统**：importance `high`
  - source_paths: pydantic_ai_slim/pydantic_ai/tools.py, pydantic_ai_slim/pydantic_ai/_function_schema.py, docs/tools.md, docs/native-tools.md, docs/api/tools.md
- **工具集与 MCP 协议**：importance `high`
  - source_paths: pydantic_ai_slim/pydantic_ai/toolsets/__init__.py, pydantic_ai_slim/pydantic_ai/toolsets/external.py, pydantic_ai_slim/pydantic_ai/toolsets/deferred_loading.py, pydantic_ai_slim/pydantic_ai/mcp.py, docs/toolsets.md
- **结构化输出**：importance `high`
  - source_paths: pydantic_ai_slim/pydantic_ai/output.py, pydantic_ai_slim/pydantic_ai/_output.py, docs/output.md, docs/api/output.md
- **多代理应用**：importance `high`
  - source_paths: docs/multi-agent-applications.md, pydantic_graph/pydantic_graph/graph.py, pydantic_graph/pydantic_graph/step.py, docs/graph.md, docs/graph/builder/index.md

## Repo Inspection Evidence / 源码检查证据

- repo_clone_verified: true
- repo_inspection_verified: true
- repo_commit: `5542eee013b1d5aaf77295ecf5f8a9b91401d52b`
- inspected_files: `pyproject.toml`, `README.md`, `uv.lock`, `docs/version-policy.md`, `docs/a2a.md`, `docs/tools.md`, `docs/AGENTS.md`, `docs/evals.md`, `docs/cli.md`, `docs/contributing.md`, `docs/index.md`, `docs/troubleshooting.md`, `docs/agent.md`, `docs/dependencies.md`, `docs/logfire.md`, `docs/agent-spec.md`, `docs/common-tools.md`, `docs/deferred-tools.md`, `docs/web.md`, `docs/hooks.md`

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：xAI: update docs and `KnownModelName` for current Grok 4.3 / 4.20 model names

- 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：xAI: update docs and `KnownModelName` for current Grok 4.3 / 4.20 model names
- 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_fecdaf4224b646929420b15d99f2a933 | https://github.com/pydantic/pydantic-ai/issues/5663 | 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：[roundtrip-sweep] ModelRequest.parts: InstructionPart fails to deserialize after serialization

- Trigger: GitHub 社区证据显示该项目存在一个配置相关的待验证问题：[roundtrip-sweep] ModelRequest.parts: InstructionPart fails to deserialize after serialization
- 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_d3718c4d2db24bceabf8d4e935091eec | https://github.com/pydantic/pydantic-ai/issues/5696 | 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 3: 来源证据：[streaming-resilience-sweep] Streaming: `stream_output()` doesn't set `is_complete=True` on early break

- Trigger: GitHub 社区证据显示该项目存在一个配置相关的待验证问题：[streaming-resilience-sweep] Streaming: `stream_output()` doesn't set `is_complete=True` on early break
- 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: 可能影响升级、迁移或版本选择。
- Evidence: community_evidence:github | cevd_9c38896c20d24789beb9604c4d4f2626 | https://github.com/pydantic/pydantic-ai/issues/5615 | 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 4: 来源证据：[aw] No-Op Runs

- 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：[aw] No-Op Runs
- 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_0ba5cb74938c4ad5aaddfd0fced08630 | https://github.com/pydantic/pydantic-ai/issues/5685 | 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: 来源证据：`MCPToolset(url, http_client=...)` crashes: factory missing `follow_redirects` kwarg passed by `FastMCP`

- 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：`MCPToolset(url, http_client=...)` crashes: factory missing `follow_redirects` kwarg passed by `FastMCP`
- 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: 可能阻塞安装或首次运行。
- Evidence: community_evidence:github | cevd_30f08a19e50a4708b984022031c55054 | https://github.com/pydantic/pydantic-ai/issues/5688 | 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 6: 失败模式：installation: v2.0.0b1 (2026-05-20)

- Trigger: Developers should check this installation risk before relying on the project: v2.0.0b1 (2026-05-20)
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: v2.0.0b1 (2026-05-20). Context: Observed when using python
- Why it matters: Upgrade or migration may change expected behavior: v2.0.0b1 (2026-05-20)
- Evidence: failure_mode_cluster:github_release | fmev_6565701c26f03fa2f9360df9e9925984 | https://github.com/pydantic/pydantic-ai/releases/tag/v2.0.0b1 | v2.0.0b1 (2026-05-20)
- 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: 失败模式：configuration: [aw] No-Op Runs

- Trigger: Developers should check this configuration risk before relying on the project: [aw] No-Op Runs
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: [aw] No-Op Runs. Context: Source discussion did not expose a precise runtime context.
- Why it matters: Developers may misconfigure credentials, environment, or host setup: [aw] No-Op Runs
- Evidence: failure_mode_cluster:github_issue | fmev_630176c03d70150693f25efc4e776f66 | https://github.com/pydantic/pydantic-ai/issues/5685 | [aw] No-Op Runs
- 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: 失败模式：configuration: `MCPToolset(url, http_client=...)` crashes: factory missing `follow_redirects` kwarg passed b...

- Trigger: Developers should check this configuration risk before relying on the project: `MCPToolset(url, http_client=...)` crashes: factory missing `follow_redirects` kwarg passed by `FastMCP`
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: `MCPToolset(url, http_client=...)` crashes: factory missing `follow_redirects` kwarg passed by `FastMCP`. Context: Observed when using python
- Why it matters: Developers may misconfigure credentials, environment, or host setup: `MCPToolset(url, http_client=...)` crashes: factory missing `follow_redirects` kwarg passed by `FastMCP`
- Evidence: failure_mode_cluster:github_issue | fmev_1c1bcf0774a223c19b0054f060350857 | https://github.com/pydantic/pydantic-ai/issues/5688 | `MCPToolset(url, http_client=...)` crashes: factory missing `follow_redirects` kwarg passed by `FastMCP`
- 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: 失败模式：configuration: v1.100.0 (2026-05-20)

- Trigger: Developers should check this configuration risk before relying on the project: v1.100.0 (2026-05-20)
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: v1.100.0 (2026-05-20). Context: Source discussion did not expose a precise runtime context.
- Why it matters: Upgrade or migration may change expected behavior: v1.100.0 (2026-05-20)
- Evidence: failure_mode_cluster:github_release | fmev_1118f6639392629ef5022ee4c6f40741 | https://github.com/pydantic/pydantic-ai/releases/tag/v1.100.0 | v1.100.0 (2026-05-20)
- 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 10: 失败模式：configuration: v1.101.0 (2026-05-21)

- Trigger: Developers should check this configuration risk before relying on the project: v1.101.0 (2026-05-21)
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: v1.101.0 (2026-05-21). Context: Source discussion did not expose a precise runtime context.
- Why it matters: Upgrade or migration may change expected behavior: v1.101.0 (2026-05-21)
- Evidence: failure_mode_cluster:github_release | fmev_e0130422b62adb912b0bfab8a362a2c7 | https://github.com/pydantic/pydantic-ai/releases/tag/v1.101.0 | v1.101.0 (2026-05-21)
- 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。