# deploy_to_apprunner - 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 deploy_to_apprunner 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_0013` 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/docs-audit/SKILL.md`, `.agents/skills/docs-planner/SKILL.md`, `.agents/skills/docs-reviewer/SKILL.md`, `.agents/skills/docs-writer/SKILL.md` Claim：`clm_0014` supported 0.86

## 它能做什么

- **Python Agent Framework**（可做安装前预览）：Core Agent class with model-driven approach, supporting simple conversational assistants to autonomous workflows via Agent(tools=[...]) pattern. 证据：`README.md`, `strands-py/AGENTS.md`, `strands-py/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**（可做安装前预览）：Model-agnostic architecture supporting Amazon Bedrock, Anthropic, OpenAI, Google Gemini, Ollama, LiteLLM, Mistral, Writer, LlamaAPI, llamacpp, and AWS SageMaker. 证据：`README.md`, `strands-py/AGENTS.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Python @tool Decorator System**（可做安装前预览）：Build tools using Python decorators with automatic docstring parsing for LLM tool descriptions, supporting hot reloading from directory. 证据：`README.md`, `README.md`, `strands-py/AGENTS.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **MCP (Model Context Protocol) Integration**（可做安装前预览）：Native MCP client supporting stdio and HTTP transports, enabling access to MCP servers for tool augmentation with background thread event loop architecture. 证据：`README.md`, `strands-py/AGENTS.md`, `strands-py/docs/MCP_CLIENT_ARCHITECTURE.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Multi-Agent Orchestration (Graph & Swarm)**（可做安装前预览）：Multi-agent patterns including Graph-based orchestration and Swarm orchestrators for complex agent coordination, with A2A protocol support. 证据：`strands-py/AGENTS.md`, `site/package.json` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Streaming Response Support**（可做安装前预览）：Real-time streaming of agent responses as they are being generated, with dedicated streaming handling in the event loop. 证据：`README.md`, `strands-py/AGENTS.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Session & Conversation Management**（可做安装前预览）：Pluggable session managers including sliding window and summarizing conversation managers for managing message history at scale. 证据：`strands-py/AGENTS.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Observability & Telemetry**（可做安装前预览）：OpenTelemetry tracing and metrics integration for agent execution monitoring and performance analysis. 证据：`strands-py/AGENTS.md`, `site/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 Output with Zod**（可做安装前预览）：Type-safe, validated responses using Zod schemas for structured output from agents, available in both Python and TypeScript SDKs. 证据：`site/docs/examples/typescript/structured_output/README.md`, `strands-py/AGENTS.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **AWS Deployment Patterns**（需要安装后验证）：CDK-based deployment examples for AWS App Runner, EC2, Fargate, Lambda, and EKS with FastAPI weather service demonstrations. 证据：`site/docs/examples/cdk/deploy_to_apprunner/README.md`, `site/docs/examples/cdk/deploy_to_ec2/README.md`, `site/docs/examples/cdk/deploy_to_fargate/README.md`, `site/docs/examples/cdk/deploy_to_lambda/README.md` 等 Claim：`clm_0010` supported 0.86
- **TypeScript/JavaScript SDK**（可做安装前预览）：Separate @strands-agents/sdk npm package for TypeScript agents with Express server deployment to Amazon Bedrock AgentCore Runtime. 证据：`site/docs/examples/typescript/deploy_to_bedrock_agentcore/README.md`, `site/package.json`, `site/docs/examples/typescript/structured_output/package.json` Claim：`clm_0005` supported 0.86, `clm_0008` supported 0.86, `clm_0011` supported 0.86
- **Agent-to-Agent (A2A) Protocol**（可做安装前预览）：Remote A2A agent client for agent-to-agent communication following the A2A protocol specification. 证据：`strands-py/AGENTS.md`, `strands-py/AGENTS.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

- `pip install strands-agents strands-agents-tools` 证据：`README.md` Claim：`clm_0015` supported 0.86
- `pip install strands-agents[bidi]` 证据：`README.md` Claim：`clm_0016` supported 0.86, `clm_0017` supported 0.86
- `pip install strands-agents[bidi,bidi-io]` 证据：`README.md` Claim：`clm_0017` supported 0.86
- `pip install hatch` 证据：`README.md` Claim：`clm_0018` 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_0013` 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/docs-audit/SKILL.md`, `.agents/skills/docs-planner/SKILL.md`, `.agents/skills/docs-reviewer/SKILL.md`, `.agents/skills/docs-writer/SKILL.md` Claim：`clm_0014` supported 0.86
- **能力存在：Python 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`, `strands-py/AGENTS.md`, `strands-py/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`, `strands-py/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：Python @tool Decorator System**（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`, `strands-py/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：MCP (Model Context Protocol) Integration**（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`, `strands-py/AGENTS.md`, `strands-py/docs/MCP_CLIENT_ARCHITECTURE.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/docs-audit/SKILL.md`, `.agents/skills/docs-planner/SKILL.md`, `.agents/skills/docs-reviewer/SKILL.md`, `.agents/skills/docs-writer/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：`README.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：`README.md`
- **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/docs-audit/SKILL.md`, `.agents/skills/docs-planner/SKILL.md`, `.agents/skills/docs-reviewer/SKILL.md`, `.agents/skills/docs-writer/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：`site/docs/examples/cdk/deploy_to_apprunner/README.md`, `site/docs/examples/cdk/deploy_to_ec2/README.md`, `site/docs/examples/cdk/deploy_to_fargate/README.md`, `site/docs/examples/cdk/deploy_to_lambda/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**：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_0019` inferred 0.45
- **命令执行会修改本地环境**：安装命令可能写入用户主目录、宿主插件目录或项目配置。 处理方式：先在隔离环境或测试账号中运行。 证据：`README.md` Claim：`clm_0020` 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

- **Python 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`, `strands-py/AGENTS.md`, `strands-py/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`, `strands-py/AGENTS.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Python @tool Decorator 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：`README.md`, `README.md`, `strands-py/AGENTS.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **MCP (Model Context Protocol) Integration**：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`, `strands-py/AGENTS.md`, `strands-py/docs/MCP_CLIENT_ARCHITECTURE.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Multi-Agent Orchestration (Graph & Swarm)**：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：`strands-py/AGENTS.md`, `site/package.json` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Streaming Response Support**：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`, `strands-py/AGENTS.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Session & Conversation Management**：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：`strands-py/AGENTS.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **Observability & Telemetry**：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：`strands-py/AGENTS.md`, `site/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 Output with Zod**：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：`site/docs/examples/typescript/structured_output/README.md`, `strands-py/AGENTS.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86 等
- **AWS Deployment Patterns**：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：`site/docs/examples/cdk/deploy_to_apprunner/README.md`, `site/docs/examples/cdk/deploy_to_ec2/README.md`, `site/docs/examples/cdk/deploy_to_fargate/README.md`, `site/docs/examples/cdk/deploy_to_lambda/README.md` 等 Claim：`clm_0010` supported 0.86
- **TypeScript/JavaScript SDK**：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：`site/docs/examples/typescript/deploy_to_bedrock_agentcore/README.md`, `site/package.json`, `site/docs/examples/typescript/structured_output/package.json` Claim：`clm_0005` supported 0.86, `clm_0008` supported 0.86, `clm_0011` supported 0.86
- **Agent-to-Agent (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：`strands-py/AGENTS.md`, `strands-py/AGENTS.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：1007
- 重要文件覆盖：40/1007
- 证据索引条目：80
- 角色 / Skill 条目：4

### 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 deploy_to_apprunner 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 deploy_to_apprunner 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 deploy_to_apprunner 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 4 个角色 / Skill / 项目文档条目。

- **docs-audit**（skill）：Assess a published or in-progress documentation page for quality, accuracy, and voice compliance. Use before rewriting a page, during periodic health checks, when community signals point to confusion, or when comparing against competitor docs. Also triggers on "audit this page", "assess the docs", "what's wrong with this page", "check docs quality", "review this doc page". 激活提示：当用户任务与“docs-audit”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/docs-audit/SKILL.md`
- **docs-planner**（skill）：Identify documentation gaps and prioritize the docs backlog. Use when planning a docs improvement sprint, after signals surface repeated friction, when new SDK features ship without docs, or for periodic health assessment. Also triggers on "plan docs work", "what docs need writing", "prioritize the backlog", "docs health check", "what should we document next". 激活提示：当用户任务与“docs-planner”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/docs-planner/SKILL.md`
- **docs-reviewer**（skill）：Review documentation drafts for voice consistency, structure, and terminology before PR submission. Use after completing a draft, when checking if docs are ready to ship, or automatically after docs-writer produces output. Also triggers on "review this draft", "check my docs", "is this ready to ship", "review before merging". 激活提示：当用户任务与“docs-reviewer”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/docs-reviewer/SKILL.md`
- **docs-writer**（skill）：Draft or rewrite Strands Agents documentation pages. Use when writing new doc pages, rewriting pages that failed audit, drafting sections for existing pages, or writing blog posts and release notes about Strands. Also triggers on "write a doc", "draft a page", "rewrite the quickstart", "add a tutorial for X", "document this feature". 激活提示：当用户任务与“docs-writer”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/docs-writer/SKILL.md`

## 证据索引

- 共索引 80 条证据。

- **AWS CDK App Runner Deployment Example**（documentation）：AWS CDK App Runner Deployment Example 证据：`site/docs/examples/cdk/deploy_to_apprunner/README.md`
- **AWS CDK EC2 Deployment Example**（documentation）：This is a TypeScript-based CDK Cloud Development Kit example that demonstrates how to deploy a Python application to AWS EC2. The example deploys a weather forecaster application that runs as a service on an EC2 instance. The application provides two weather endpoints: 证据：`site/docs/examples/cdk/deploy_to_ec2/README.md`
- **AWS CDK Fargate Deployment Example**（documentation）：This is a TypeScript-based CDK Cloud Development Kit example that demonstrates how to deploy a Python application to AWS Fargate. The example deploys a weather forecaster application that runs as a containerized service in AWS Fargate with an Application Load Balancer. The application is built with FastAPI and provides two weather endpoints: 证据：`site/docs/examples/cdk/deploy_to_fargate/README.md`
- **AWS CDK Lambda Deployment Example**（documentation）：This is a TypeScript-based CDK Cloud Development Kit example that demonstrates how to deploy a Python function to AWS Lambda. The example deploys a weather forecaster application that requires AWS authentication to invoke the Lambda function. 证据：`site/docs/examples/cdk/deploy_to_lambda/README.md`
- **Amazon EKS Deployment Example**（documentation）：This is an example that demonstrates how to deploy a Python application to Amazon EKS. The example deploys a weather forecaster application that runs as a containerized service in Amazon EKS with an Application Load Balancer. The application is built with FastAPI and provides two weather endpoints: 证据：`site/docs/examples/deploy_to_eks/README.md`
- **TypeScript Agent Deployment to Amazon Bedrock AgentCore Runtime**（documentation）：TypeScript Agent Deployment to Amazon Bedrock AgentCore Runtime 证据：`site/docs/examples/typescript/deploy_to_bedrock_agentcore/README.md`
- **Structured Output Example**（documentation）：Demonstrates how to use structured output with Strands Agents to get type-safe, validated responses using Zod schemas. 证据：`site/docs/examples/typescript/structured_output/README.md`
- **Developer Documentation**（documentation）：This folder contains documentation for contributors and developers working on the Strands Agents SDK. 证据：`strands-py/docs/README.md`
- **Feature Overview**（documentation）：A model-driven approach to building AI agents in just a few lines of code. 证据：`README.md`
- **Design Documents**（documentation）：This folder contains design documents for significant features and changes to Strands Agents. These documents capture the context, decision, and consequences of architectural choices. 证据：`designs/README.md`
- **Team Documentation**（documentation）：This folder contains internal documentation about how the Strands team builds and maintains the SDK. These documents help contributors understand our design philosophy and decision-making process. 证据：`site/team/README.md`
- **Agent Development Guide - strands-agents/private-docs-staging**（documentation）：Agent Development Guide - strands-agents/private-docs-staging 证据：`site/AGENTS.md`
- **AGENTS.md**（documentation）：This document provides context, patterns, and guidelines for AI coding assistants working in this repository. For human contributors, see CONTRIBUTING.md ../CONTRIBUTING.md . 证据：`strands-py/AGENTS.md`
- **MCP Client Architecture**（documentation）：The MCPClient enables developers to use MCP tools in Strands agents without dealing with async complexity. Since MCP requires async operations but Strands aims for simple synchronous usage agent = Agent ; agent "Do something" , the client uses a background thread with its own event loop to handle MCP communication. This creates challenges around thread synchronization, hanging prevention, and connection stability that this architecture addresses. 证据：`strands-py/docs/MCP_CLIENT_ARCHITECTURE.md`
- **Package**（package_manifest）：{ "name": "docs", "version": "1.0.0", "type": "module", "scripts": { "build": "astro build", "build:all": "npm run sdk:clone && npm run sdk:generate && npm run build", "dev": "astro dev", "test": "vitest run", "typecheck": "tsc --noEmit", "typecheck:snippets": "npm run typecheck --prefix test-snippets", "format": "prettier --write docs 'src/content/docs/ / .ts'", "format:check": "prettier --check docs 'src/content/docs/ / .ts'", "clean": "rm -rf .build && rm -rf .astro", "preview": "astro preview", "cms:build": "npm run build:all", "sdk:clone": "tsx scripts/clone-sdks.ts", "sdk:generate:py": "command -v uv /dev/null 2 &1 && uv run scripts/api-generation-python.py pip install pydoc-markdown… 证据：`site/package.json`
- **Contributing Guidelines**（documentation）：Thank you for your interest in contributing to our project. Whether it's a bug report, new feature, correction, or additional documentation, we greatly value feedback and contributions from our community. 证据：`CONTRIBUTING.md`
- **Contributing Guidelines**（documentation）：Thank you for your interest in contributing to our project. Whether it's a bug report, new feature, correction, or additional documentation, we greatly value feedback and contributions from our community. 证据：`site/CONTRIBUTING.md`
- **Package**（package_manifest）：{ "name": "deploy to apprunner", "version": "0.1.0", "description": "CDK TypeScript project to deploy a sample Agent on AWS App Runner", "private": true, "bin": { "cdk-app": "bin/cdk-app.js" }, "scripts": { "format": "prettier --write .", "watch": "tsc -w", "test": "vitest run", "cdk": "cdk" }, "devDependencies": { "@types/node": "22.15.3", "aws-cdk": "2.1012.0", "prettier": "~3.5.3", "tsx": "^4.7.0", "typescript": "~5.8.3", "vitest": "^3.1.2" }, "dependencies": { "aws-cdk-lib": "2.192.0", "constructs": "^10.2.0" }, "prettier": { "printWidth": 120 } } 证据：`site/docs/examples/cdk/deploy_to_apprunner/package.json`
- **Package**（package_manifest）：{ "name": "deploy to ec2", "version": "0.1.0", "description": "CDK TypeScript project to deploy a sample agent to EC2", "private": true, "bin": { "cdk-app": "bin/cdk-app.js" }, "scripts": { "format": "prettier --write .", "watch": "tsc -w", "test": "vitest run", "cdk": "cdk" }, "devDependencies": { "@types/node": "22.15.3", "aws-cdk": "2.1012.0", "prettier": "~3.5.3", "tsx": "^4.7.0", "typescript": "~5.8.3", "vitest": "^3.1.2" }, "dependencies": { "aws-cdk-lib": "2.192.0", "constructs": "^10.2.0" }, "prettier": { "printWidth": 120 } } 证据：`site/docs/examples/cdk/deploy_to_ec2/package.json`
- **Package**（package_manifest）：{ "name": "deploy to lambda", "version": "0.1.0", "description": "CDK TypeScript project to deploy a sample agent Lambda function", "private": true, "bin": { "cdk-app": "bin/cdk-app.js" }, "scripts": { "format": "prettier --write .", "watch": "tsc -w", "test": "vitest run", "cdk": "cdk" }, "devDependencies": { "@types/node": "22.15.3", "aws-cdk": "2.1012.0", "prettier": "~3.5.3", "tsx": "^4.7.0", "typescript": "~5.8.3", "vitest": "^3.1.2" }, "dependencies": { "aws-cdk-lib": "2.192.0", "constructs": "^10.2.0" }, "prettier": { "printWidth": 120 } } 证据：`site/docs/examples/cdk/deploy_to_fargate/package.json`
- **Package**（package_manifest）：{ "name": "deploy to lambda", "version": "0.1.0", "description": "CDK TypeScript project to deploy a sample Agent Lambda function", "private": true, "bin": { "cdk-app": "bin/cdk-app.js" }, "scripts": { "format": "prettier --write .", "watch": "tsc -w", "test": "vitest run", "cdk": "cdk" }, "devDependencies": { "@types/node": "22.15.3", "aws-cdk": "2.1012.0", "prettier": "~3.5.3", "tsx": "^4.7.0", "typescript": "~5.8.3", "vitest": "^3.1.2" }, "dependencies": { "aws-cdk-lib": "2.192.0", "constructs": "^10.2.0" }, "prettier": { "printWidth": 120 } } 证据：`site/docs/examples/cdk/deploy_to_lambda/package.json`
- **Package**（package_manifest）：{ "name": "bedrock-agentcore-typescript-example", "version": "1.0.0", "description": "Example TypeScript agent deployed to Amazon Bedrock AgentCore Runtime", "type": "module", "scripts": { "build": "tsc", "start": "node dist/index.js", "dev": "tsc && node dist/index.js", "test:invoke": "npx tsx invoke.ts" }, "dependencies": { "@strands-agents/sdk": "latest", "@aws-sdk/client-bedrock-agentcore": "latest", "express": "^4.18.2", "zod": "^4.1.12" }, "devDependencies": { "@types/express": "^4.17.21", "typescript": "^5.3.3" } } 证据：`site/docs/examples/typescript/deploy_to_bedrock_agentcore/package.json`
- **Package**（package_manifest）：{ "name": "structured-output-example", "version": "1.0.0", "description": "Structured output example using Strands Agents TypeScript SDK with Zod schemas", "type": "module", "scripts": { "start": "npx tsx structured output.ts" }, "dependencies": { "@strands-agents/sdk": "latest", "zod": "^4.1.12" }, "devDependencies": { "tsx": "^4.21.0", "typescript": "^5.3.3" } } 证据：`site/docs/examples/typescript/structured_output/package.json`
- **Documentation Audit**（skill_instruction）：Assess an existing documentation page. Produces a structured assessment with specific improvement recommendations. 证据：`.agents/skills/docs-audit/SKILL.md`
- **Documentation Planner**（skill_instruction）：Identify gaps in the docs site and produce a prioritized backlog. 证据：`.agents/skills/docs-planner/SKILL.md`
- **Documentation Reviewer**（skill_instruction）：Scope: Voice, style, structure, and terminology of drafts in progress. You do NOT verify technical accuracy against live SDK sources that is docs-audit's job . 证据：`.agents/skills/docs-reviewer/SKILL.md`
- **Documentation Writer**（skill_instruction）：Draft or rewrite Strands Agents documentation following the five-layer voice stack. 证据：`.agents/skills/docs-writer/SKILL.md`
- **Package**（package_manifest）：{ "name": "test-snippets", "version": "1.0.0", "private": true, "type": "module", "scripts": { "typecheck": "tsc --noEmit" }, "dependencies": { "@strands-agents/sdk": "file:../.build/sdk-typescript/strands-ts", "@aws-sdk/client-s3": "^3.943.0", "@modelcontextprotocol/sdk": "^1.0.0", "@opentelemetry/sdk-trace-node": "^2.6.1", "express": "^5.1.0", "zod": "^4.1.12" }, "devDependencies": { "@types/express": "^5.0.5", "@types/node": "^24.10.1", "typescript": "^5.9.3" } } 证据：`site/test-snippets/package.json`
- **License**（source_file）：Apache License Version 2.0, January 2004 http://www.apache.org/licenses/ 证据：`LICENSE`
- **License**（source_file）：Apache License Version 2.0, January 2004 http://www.apache.org/licenses/ 证据：`site/LICENSE`
- **Multi-Agent Example**（documentation）：This directory contains the implementation files for the Multi-Agent Example architecture, where specialized agents work together under the coordination of a central orchestrator. 证据：`site/docs/examples/python/multi_agent_example/index.md`
- **Hooks System**（documentation）：The hooks system enables extensible agent functionality through strongly-typed event callbacks. 证据：`strands-py/docs/HOOKS.md`
- **Pull Request Description Guidelines**（documentation）：Pull Request Description Guidelines 证据：`strands-py/docs/PR.md`
- **Style Guide**（documentation）：The Strands Agents style guide aims to establish consistent formatting, naming conventions, and structure across all code in the repository. We strive to make our code clean, readable, and maintainable. 证据：`strands-py/docs/STYLE_GUIDE.md`
- **Include required package dependencies for mypy**（source_file）：build-system requires = "hatchling", "hatch-vcs" build-backend = "hatchling.build" 证据：`strands-py/pyproject.toml`
- **Evaluate Graph**（source_file）：async def async graph example ⋮---- def research graph case: Case ⋮---- researcher = Agent name="researcher", system prompt="You are a research specialist..." analyst = Agent name="analyst", system prompt="You are a data analysis specialist..." fact checker = Agent name="fact checker", system prompt="You are a fact checking specialist..." report writer = Agent name="report writer", system prompt="You are a report writing specialist..." ⋮---- builder = GraphBuilder ⋮---- graph = builder.build ⋮---- result = graph case.input interactions = graph extractor.extract graph interactions result ⋮---- test1 = Case test2 = Case input="Research the impact of robotics on healthcare and create a short r… 证据：`site/docs/examples/evals-sdk/evaluate_graph.py`
- **Evaluate Swarm**（source_file）：async def async swarm example ⋮---- def sde swarm case: Case ⋮---- researcher = Agent name="researcher", system prompt="You are a research specialist...", callback handler=None coder = Agent name="coder", system prompt="You are a coding specialist...", callback handler=None reviewer = Agent name="reviewer", system prompt="You are a code review specialist...", callback handler=None architect = Agent ⋮---- swarm = Swarm ⋮---- result = swarm case.input interaction info = swarm extractor.extract swarm interactions result ⋮---- test1 = Case ⋮---- interaction evaluator = InteractionsEvaluator trajectory evaluator = TrajectoryEvaluator ⋮---- trajectory experiment = Experiment cases= test1 , evalua… 证据：`site/docs/examples/evals-sdk/evaluate_swarm.py`
- **System prompt to determine action**（source_file）：DEFAULT KB ID = "demokb123" KB ID = os.environ.get "STRANDS KNOWLEDGE BASE ID" ⋮---- KB ID = DEFAULT KB ID ⋮---- System prompt to determine action ACTION SYSTEM PROMPT = """ result = agent.tool.use llm ⋮---- Clean and extract the action action text = str result .lower .strip ⋮---- Default to retrieve if response isn't clear ⋮---- def run kb agent query ⋮---- """Process a user query with the knowledge base agent.""" agent = Agent tools= memory, use llm ⋮---- Determine the action - store or retrieve action = determine action agent, query ⋮---- result = agent.tool.memory ⋮---- result str = str result ⋮---- answer = agent.tool.use llm ⋮---- user input = input "\n " 证据：`site/docs/examples/python/knowledge_base_agent.py`
- **Memory Agent**（source_file）：logger = logging.getLogger name ⋮---- USER ID = "mem0 user" ⋮---- MEMORY SYSTEM PROMPT = f"""You are a personal assistant that maintains context by remembering user details. ⋮---- memory agent = Agent ⋮---- def initialize demo memories ⋮---- content = """My name is Alex. I like to travel and stay in Airbnbs rather than hotels. I am planning a trip to Japan next spring. I enjoy hiking and outdoor photography as hobbies. I have a dog named Max. My favorite cuisine is Italian food.""" ⋮---- user input = input "\n " 证据：`site/docs/examples/python/memory_agent.py`
- **Navigation**（source_file）：navbar: - label: Home href: / - label: Docs href: /docs/user-guide/quickstart/typescript/ basePath: /docs/user-guide/ - label: Examples href: /docs/examples/ basePath: /docs/examples/ - label: Community href: /docs/community/community-packages/ basePath: - /docs/community/ - /docs/labs/ - /docs/contribute/ - label: API Reference href: /docs/api/typescript/ basePath: /docs/api/ - label: Blog href: /blog/ basePath: /blog/ align: right - label: Labs href: https://github.com/strands-labs external: true align: right sidebar: - label: Docs items: - docs/index - label: Get Started items: - docs/user-guide/quickstart/overview - label: "Quickstart: Python" slug: docs/user-guide/quickstart/python - l… 证据：`site/src/config/navigation.yml`
- **Getting Started**（source_file）：A collection of sample implementations to help you get started with Strands Agents. From simple agents to complex multi-agent systems, each example illustrates key concepts and patterns you can adapt for your own projects. 证据：`site/src/content/docs/examples/README.mdx`
- **Agent**（source_file）：app = FastAPI title="Strands Agent Server", version="1.0.0" ⋮---- model = OpenAIModel model id="gpt-4o" ⋮---- strands agent = Agent model=model ⋮---- class InvocationRequest BaseModel ⋮---- input: Dict str, Any ⋮---- class InvocationResponse BaseModel ⋮---- output: Dict str, Any ⋮---- @app.post "/invocations", response model=InvocationResponse async def invoke agent request: InvocationRequest ⋮---- user message = request.input.get "prompt", "" ⋮---- result = strands agent user message response = { ⋮---- @app.get "/ping" async def ping ⋮---- def main 证据：`site/src/content/docs/user-guide/deploy/deploy_to_docker/agent.py`
- **Init**（source_file）：all = 证据：`strands-py/src/strands/__init__.py`
- **Init**（source_file）：all = ⋮---- def getattr name: str - Any 证据：`strands-py/src/strands/agent/__init__.py`
- **A2A Agent**（source_file）：logger = logging.getLogger name ⋮---- DEFAULT TIMEOUT = 300 ⋮---- TERMINAL STATES = {state for state, reason in STATE TO STOP REASON.items if reason == "end turn"} INPUT STATES = {state for state, reason in STATE TO STOP REASON.items if reason == "interrupt"} COMPLETE STATES = TERMINAL STATES INPUT STATES ⋮---- class A2AAgent AgentBase ⋮---- result: AgentResult None = None ⋮---- result = event "result" ⋮---- last event = None last complete event = None ⋮---- last event = event ⋮---- last complete event = event ⋮---- final event = last complete event or last event ⋮---- result = convert response to agent result final event ⋮---- async def get agent card self - AgentCard ⋮---- resolver = A2AC… 证据：`strands-py/src/strands/agent/a2a_agent.py`
- **Register built-in plugins**（source_file）：logger = logging.getLogger name ⋮---- T = TypeVar "T", bound=BaseModel ⋮---- class DefaultCallbackHandlerSentinel ⋮---- class DefaultRetryStrategySentinel ⋮---- DEFAULT CALLBACK HANDLER = DefaultCallbackHandlerSentinel DEFAULT RETRY STRATEGY = DefaultRetryStrategySentinel DEFAULT AGENT NAME = "Strands Agents" DEFAULT AGENT ID = "default" ⋮---- class Agent AgentBase ⋮---- ToolCaller = ToolCaller ⋮---- Register built-in plugins ⋮---- def cancel self - None ⋮---- """Cancel the currently running agent invocation. This method is thread-safe and can be called from any context e.g., another thread, web request handler, background task . The agent will stop gracefully at the next checkpoint: - Duri… 证据：`strands-py/src/strands/agent/agent.py`
- **Agent Result**（source_file）：@dataclass class AgentResult ⋮---- stop reason: StopReason message: Message metrics: EventLoopMetrics state: Any interrupts: Sequence Interrupt None = None structured output: BaseModel None = None ⋮---- @property def context size self - int None ⋮---- @property def projected context size self - int None ⋮---- def str self - str ⋮---- content array = self.message.get "content", result = "" ⋮---- citations block = item "citationsContent" ⋮---- @classmethod def from dict cls, data: dict str, Any - "AgentResult" ⋮---- message = cast Message, data.get "message" stop reason = cast StopReason, data.get "stop reason" ⋮---- def to dict self - dict str, Any 证据：`strands-py/src/strands/agent/agent_result.py`
- **Base**（source_file）：@runtime checkable class AgentBase Protocol 证据：`strands-py/src/strands/agent/base.py`
- **Init**（source_file）：all = 证据：`strands-py/src/strands/agent/conversation_manager/__init__.py`
- **Always subscribe — the threshold check happens inside the handler**（source_file）：logger = logging.getLogger name ⋮---- DEFAULT COMPRESSION THRESHOLD = 0.7 DEFAULT CONTEXT WINDOW LIMIT = 200 000 ⋮---- class ProactiveCompressionConfig TypedDict, total=False ⋮---- compression threshold: float ⋮---- class ConversationManager ABC, HookProvider ⋮---- def init self, , proactive compression: Union bool, "ProactiveCompressionConfig", None = None - None ⋮---- threshold: float None = DEFAULT COMPRESSION THRESHOLD ⋮---- threshold = proactive compression.get "compression threshold", DEFAULT COMPRESSION THRESHOLD ⋮---- threshold = None ⋮---- def register hooks self, registry: HookRegistry, kwargs: Any - None ⋮---- """Register hooks for agent lifecycle events. Always registers a Befor… 证据：`strands-py/src/strands/agent/conversation_manager/conversation_manager.py`
- **Null Conversation Manager**（source_file）：class NullConversationManager ConversationManager ⋮---- def apply management self, agent: "Agent", kwargs: Any - None ⋮---- def reduce context self, agent: "Agent", e: Exception None = None, kwargs: Any - None 证据：`strands-py/src/strands/agent/conversation_manager/null_conversation_manager.py`
- **Sliding Window Conversation Manager**（source_file）：logger = logging.getLogger name ⋮---- PRESERVE CHARS = 200 ⋮---- class SlidingWindowConversationManager ConversationManager ⋮---- def register hooks self, registry: "HookRegistry", kwargs: Any - None ⋮---- def on before model call self, event: BeforeModelCallEvent - None ⋮---- should apply = False ⋮---- should apply = True ⋮---- should apply = self. model call count % self.per turn == 0 ⋮---- def get state self - dict str, Any ⋮---- state = super .get state ⋮---- def restore from session self, state: dict str, Any - list None ⋮---- result = super .restore from session state ⋮---- def apply management self, agent: "Agent", kwargs: Any - None ⋮---- messages = agent.messages ⋮---- def reduce c… 证据：`strands-py/src/strands/agent/conversation_manager/sliding_window_conversation_manager.py`
- **Summarizing Conversation Manager**（source_file）：logger = logging.getLogger name ⋮---- DEFAULT SUMMARIZATION PROMPT = """You are a conversation summarizer. Provide a concise summary of the conversation \ ⋮---- class SummarizingConversationManager ConversationManager ⋮---- @override def restore from session self, state: dict str, Any - list Message None ⋮---- def get state self - dict str, Any ⋮---- def apply management self, agent: "Agent", kwargs: Any - None ⋮---- def reduce context self, agent: "Agent", e: Exception None = None, kwargs: Any - None ⋮---- def summarize oldest self, agent: "Agent" - None ⋮---- messages to summarize count = max 1, int len agent.messages self.summary ratio ⋮---- messages to summarize count = min ⋮---- messag… 证据：`strands-py/src/strands/agent/conversation_manager/summarizing_conversation_manager.py`
- **State**（source_file）：AgentState = JSONSerializableDict 证据：`strands-py/src/strands/agent/state.py`
- **Init**（source_file）：all = "event loop" 证据：`strands-py/src/strands/event_loop/__init__.py`
- **Event Loop**（source_file）：logger = logging.getLogger name ⋮---- MAX ATTEMPTS = 6 INITIAL DELAY = 4 MAX DELAY = 240 ⋮---- def has tool use in latest message messages: "Messages" - bool ⋮---- latest message = messages -1 content blocks = latest message.get "content", ⋮---- async def estimate input tokens agent: "Agent" - int ⋮---- messages = agent.messages ⋮---- last assistant idx = -1 ⋮---- last assistant idx = i ⋮---- usage = messages last assistant idx "metadata" "usage" known baseline = usage "inputTokens" + usage "outputTokens" new messages = messages last assistant idx + 1 : ⋮---- tool specs = agent.tool registry.get all tool specs ⋮---- structured output context = structured output context or StructuredOutputCo… 证据：`strands-py/src/strands/event_loop/event_loop.py`
- **Init**（source_file）：all = "checkpoint", "config to agent", "tools", "steering" 证据：`strands-py/src/strands/experimental/__init__.py`
- **Prepare Agent constructor arguments**（source_file）：AGENT CONFIG SCHEMA = { ⋮---- VALIDATOR = jsonschema.Draft7Validator AGENT CONFIG SCHEMA ⋮---- def config to agent config: str dict str, Any , kwargs: dict str, Any - Any ⋮---- file path = config ⋮---- file path = file path 7: ⋮---- config path = Path file path ⋮---- config dict = json.load f ⋮---- config dict = config.copy ⋮---- error path = " - ".join str p for p in e.absolute path if e.absolute path else "root" ⋮---- Prepare Agent constructor arguments agent kwargs = {} ⋮---- Map configuration keys to Agent constructor parameters config mapping = { 证据：`strands-py/src/strands/experimental/agent_config.py`
- **Init**（source_file）：all = ⋮---- def getattr name: str - Any 证据：`strands-py/src/strands/experimental/bidi/__init__.py`
- **Init**（source_file）：all = " TaskGroup", " TaskPool" ⋮---- async def stop all funcs: Callable ..., Awaitable None - None ⋮---- exceptions = 证据：`strands-py/src/strands/experimental/bidi/_async/__init__.py`
- 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：`site/docs/examples/cdk/deploy_to_apprunner/README.md`, `site/docs/examples/cdk/deploy_to_ec2/README.md`, `site/docs/examples/cdk/deploy_to_fargate/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：`site/docs/examples/cdk/deploy_to_apprunner/README.md`, `site/docs/examples/cdk/deploy_to_ec2/README.md`, `site/docs/examples/cdk/deploy_to_fargate/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。

- **项目概述**：importance `high`
  - source_paths: README.md, strands-py/src/strands/__init__.py, strands-py/src/strands/agent/__init__.py
- **快速开始**：importance `high`
  - source_paths: strands-py/pyproject.toml, strands-py/src/strands/agent/agent.py, strands-py/src/strands/models/_defaults.py
- **Agent 系统架构**：importance `high`
  - source_paths: strands-py/src/strands/agent/agent.py, strands-py/src/strands/agent/base.py, strands-py/src/strands/agent/state.py, strands-py/src/strands/agent/agent_result.py, strands-py/src/strands/hooks/events.py
- **模型提供商**：importance `high`
  - source_paths: strands-py/src/strands/models/model.py, strands-py/src/strands/models/bedrock.py, strands-py/src/strands/models/anthropic.py, strands-py/src/strands/models/gemini.py, strands-py/src/strands/models/openai.py
- **工具系统**：importance `high`
  - source_paths: strands-py/src/strands/tools/decorator.py, strands-py/src/strands/tools/registry.py, strands-py/src/strands/tools/tools.py, strands-py/src/strands/tools/_tool_helpers.py, strands-py/src/strands/tools/executors/_executor.py
- **会话与上下文管理**：importance `medium`
  - source_paths: strands-py/src/strands/agent/conversation_manager/conversation_manager.py, strands-py/src/strands/agent/conversation_manager/sliding_window_conversation_manager.py, strands-py/src/strands/agent/conversation_manager/summarizing_conversation_manager.py, strands-py/src/strands/session/session_manager.py, strands-py/src/strands/session/file_session_manager.py
- **多 Agent 系统**：importance `high`
  - source_paths: strands-py/src/strands/multiagent/__init__.py, strands-py/src/strands/multiagent/graph.py, strands-py/src/strands/multiagent/swarm.py, strands-py/src/strands/multiagent/a2a/__init__.py, strands-py/src/strands/multiagent/a2a/server.py
- **MCP 集成**：importance `high`
  - source_paths: strands-py/src/strands/tools/mcp/__init__.py, strands-py/src/strands/tools/mcp/mcp_client.py, strands-py/src/strands/tools/mcp/mcp_agent_tool.py, strands-py/src/strands/tools/mcp/mcp_tasks.py, strands-py/src/strands/tools/mcp/mcp_types.py

## Repo Inspection Evidence / 源码检查证据

- repo_clone_verified: true
- repo_inspection_verified: true
- repo_commit: `de0e9508aa1bf72ed8e5113d19b0d282329d94ff`
- inspected_files: `README.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: 仓库名和安装名不一致

- Trigger: 仓库名 `sdk-python` 与安装入口 `strands-agents` 不完全一致。
- Host AI rule: 在 npm/PyPI/GitHub 上确认包名映射和官方 README 说明。
- Why it matters: 用户照着仓库名搜索包或照着包名找仓库时容易走错入口。
- Evidence: identity.distribution | github_repo:983715534 | https://github.com/strands-agents/sdk-python | repo=sdk-python; install=strands-agents
- 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: 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:983715534 | https://github.com/strands-agents/sdk-python | 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 3: 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:983715534 | https://github.com/strands-agents/sdk-python | 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 4: 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:983715534 | https://github.com/strands-agents/sdk-python | 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 5: 存在安全注意事项

- Trigger: No sandbox install has been executed yet; downstream must verify before user use.
- Host AI rule: 转成明确权限清单和安全审查提示。
- Why it matters: 用户安装前需要知道权限边界和敏感操作。
- Evidence: risks.safety_notes | github_repo:983715534 | https://github.com/strands-agents/sdk-python | No sandbox install has been executed yet; downstream must verify before user use.
- 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:983715534 | https://github.com/strands-agents/sdk-python | 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:983715534 | https://github.com/strands-agents/sdk-python | 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:983715534 | https://github.com/strands-agents/sdk-python | 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。