# prefect - 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 prefect 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

- **希望把专业流程带进宿主 AI 的用户**：仓库包含 Skill 文档。 证据：`.claude/skills/agents-md-sync/SKILL.md`, `.claude/skills/backlog-management/SKILL.md`, `.claude/skills/document-changes/SKILL.md`, `.claude/skills/write-docs/SKILL.md` Claim：`clm_0003` supported 0.86

## 它能做什么

- **AI Skill / Agent 指令资产库**（可做安装前预览）：项目包含可被宿主 AI 读取的 Skill 或 Agent 指令文件，可用于把专业流程带入 Claude、Codex、Cursor 等宿主。 证据：`.claude/skills/agents-md-sync/SKILL.md`, `.claude/skills/backlog-management/SKILL.md`, `.claude/skills/document-changes/SKILL.md`, `.claude/skills/write-docs/SKILL.md` Claim：`clm_0001` supported 0.86
- **命令行启动或安装流程**（需要安装后验证）：项目文档中存在可执行命令，真实使用需要在本地或宿主环境中运行这些命令。 证据：`README.md` Claim：`clm_0002` supported 0.86

## 怎么开始

- `pip install -U prefect` 证据：`README.md` Claim：`clm_0004` supported 0.86

## 继续前判断卡

- **当前建议**：仅建议沙盒试装
- **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**：仅建议沙盒试装
- **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**：真实输出质量不能在安装前相信。
- **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：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：`.claude/skills/agents-md-sync/SKILL.md`, `.claude/skills/backlog-management/SKILL.md`, `.claude/skills/document-changes/SKILL.md`, `.claude/skills/write-docs/SKILL.md` Claim：`clm_0003` supported 0.86
- **能力存在：AI Skill / Agent 指令资产库**（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：`.claude/skills/agents-md-sync/SKILL.md`, `.claude/skills/backlog-management/SKILL.md`, `.claude/skills/document-changes/SKILL.md`, `.claude/skills/write-docs/SKILL.md` Claim：`clm_0001` supported 0.86
- **能力存在：命令行启动或安装流程**（supported）：可以相信项目包含这类能力线索；是否适合你的具体任务仍要试用或安装后验证。 证据：`README.md` Claim：`clm_0002` supported 0.86
- **存在 Quick Start / 安装命令线索**（supported）：可以相信项目文档出现过启动或安装入口；不要因此直接在主力环境运行。 证据：`README.md` Claim：`clm_0004` supported 0.86

### 现在还不能相信

- **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：`.claude/skills/agents-md-sync/SKILL.md`, `.claude/skills/backlog-management/SKILL.md`, `.claude/skills/document-changes/SKILL.md`, `.claude/skills/write-docs/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：`.claude/skills/agents-md-sync/SKILL.md`, `.claude/skills/backlog-management/SKILL.md`, `.claude/skills/document-changes/SKILL.md`, `.claude/skills/write-docs/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`
- **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_0005` 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_0006` supported 0.86
- **待确认**：真实安装后是否与用户当前宿主 AI 版本兼容？。原因：兼容性只能通过实际宿主环境验证。
- **待确认**：项目输出质量是否满足用户具体任务？。原因：安装前预览只能展示流程和边界，不能替代真实评测。
- **待确认**：安装命令是否需要网络、权限或全局写入？。原因：这影响企业环境和个人环境的安装风险。

## 开工前工作上下文

### 加载顺序

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

### 任务路由

- **AI Skill / Agent 指令资产库**：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：`.claude/skills/agents-md-sync/SKILL.md`, `.claude/skills/backlog-management/SKILL.md`, `.claude/skills/document-changes/SKILL.md`, `.claude/skills/write-docs/SKILL.md` Claim：`clm_0001` supported 0.86
- **命令行启动或安装流程**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`README.md` Claim：`clm_0002` supported 0.86

### 上下文规模

- 文件总数：4430
- 重要文件覆盖：40/4430
- 证据索引条目：79
- 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 prefect 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 prefect 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 prefect 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。

- **agents-md-sync**（skill）：Analyze code changes on the current branch and recommend updates to AGENTS.md files that have become stale. Use this skill before creating a PR, when the user asks to check AGENTS.md freshness, or when preparing an AGENTS.md update after a PR merges to main. Trigger whenever the user mentions syncing, updating, or checking AGENTS.md files, or when they're about to create a PR that touches code structure new modules,… 激活提示：当用户任务与“agents-md-sync”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/agents-md-sync/SKILL.md`
- **backlog-management**（skill）：Manage open-source GitHub issue backlogs especially PrefectHQ/prefect in batches. Use for backlog cleanup days, stale review, duplicate pruning, missing-info follow-up, and prioritized queue curation across many open issues. 激活提示：当用户任务与“backlog-management”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/backlog-management/SKILL.md`
- **document-changes**（skill）：Analyze source code changes on the current branch and identify documentation that is stale, inaccurate, or missing. Use before creating a PR to check if docs need updating, or in CI after merges to main to automatically fix docs. Trigger when the user mentions "doc changes", "docs update", "check docs", "document changes", or wants to know if code changes affect documentation. 激活提示：当用户任务与“document-changes”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/document-changes/SKILL.md`
- **write-docs**（skill）：Comprehensive guide for writing and updating Prefect documentation. Use when creating new doc pages, updating existing docs, or working with Mintlify components and code example testing. 激活提示：当用户任务与“write-docs”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/write-docs/SKILL.md`

## 证据索引

- 共索引 79 条证据。

- **Prefect**（documentation）：Installation · Quickstart · Build workflows · Deploy workflows · Prefect Cloud 证据：`README.md`
- **Benchmarks**（documentation）：To run a subset of benchmarks, use python benches e.g. python benches bench flows.py 证据：`benches/README.md`
- **prefect-client**（documentation）：The prefect-client package is a minimal-installation of prefect which is designed for interacting with Prefect Cloud or remote any prefect server. It sheds some functionality and dependencies in exchange for a smaller installation size, making it ideal for use in lightweight or ephemeral environments. These characteristics make it ideal for use in lambdas or other resource-constrained environments. 证据：`client/README.md`
- **Prefect Documentation**（documentation）：Structural context for any agent working in docs/ . For the full writing guide page types, components, code testing, style , use the /write-docs skill. 证据：`docs/AGENTS.md`
- **investigating server performance**（documentation）：allow the following scripts to run via chmod +x or similar 证据：`load_testing/README.md`
- **Prefect UI**（documentation）：Compiles and hot-reloads for development 证据：`ui-v2/README.md`
- **Prefect UI**（documentation）：Compiles and hot-reloads for development 证据：`ui/README.md`
- **local-telemetry**（documentation）：This directory includes an OpenTelemetry stack that you can run locally for testing and debugging. 证据：`load_testing/local-telemetry/README.md`
- **Readme**（documentation）：This is a simple test to ensure we can make a websocket connection through a proxy server. It sets up a simple server and a squid proxy server. The proxy server is inaccessible from the host machine, so we can confirm the proxy connection is working. 证据：`scripts/proxy-test/README.md`
- **prefect-aws**（documentation）：Build production-ready data workflows that seamlessly integrate with AWS services. prefect-aws provides battle-tested blocks, tasks, and infrastructure integrations for AWS, featuring support for ECS, S3, Secrets Manager, Lambda, Batch, and Glue. 证据：`src/integrations/prefect-aws/README.md`
- **prefect-azure**（documentation）：prefect-azure is a collection of Prefect integrations for orchestration workflows with Azure. 证据：`src/integrations/prefect-azure/README.md`
- **prefect-bitbucket**（documentation）：See the docs at https://docs.prefect.io/integrations/prefect-bitbucket https://docs.prefect.io for more information. 证据：`src/integrations/prefect-bitbucket/README.md`
- **prefect-dask**（documentation）：See the docs at https://docs.prefect.io/integrations/prefect-dask https://docs.prefect.io/integrations/prefect-dask for more information. 证据：`src/integrations/prefect-dask/README.md`
- **prefect-databricks**（documentation）：See the docs at https://docs.prefect.io/integrations/prefect-databricks https://docs.prefect.io/integrations/prefect-databricks for more information. 证据：`src/integrations/prefect-databricks/README.md`
- **prefect-dbt**（documentation）：See the docs at https://docs.prefect.io/integrations/prefect-dbt https://docs.prefect.io/integrations/prefect-dbt for more information. 证据：`src/integrations/prefect-dbt/README.md`
- **prefect-docker**（documentation）：See the docs at https://docs.prefect.io/integrations/prefect-docker https://docs.prefect.io/integrations/prefect-docker for more information. 证据：`src/integrations/prefect-docker/README.md`
- **prefect-email**（documentation）：See the docs at https://docs.prefect.io/integrations/prefect-email https://docs.prefect.io/integrations/prefect-email for more information. 证据：`src/integrations/prefect-email/README.md`
- **prefect-gcp**（documentation）：prefect-gcp makes it easy to leverage the capabilities of Google Cloud Platform GCP in your flows, featuring support for Vertex AI, Cloud Run, BigQuery, Cloud Storage, and Secret Manager. 证据：`src/integrations/prefect-gcp/README.md`
- **prefect-github**（documentation）：See the docs at https://docs.prefect.io/integrations/prefect-github https://docs.prefect.io/integrations/prefect-github for more information. 证据：`src/integrations/prefect-github/README.md`
- **prefect-gitlab**（documentation）：See the docs at https://docs.prefect.io/integrations/prefect-gitlab https://docs.prefect.io/integrations/prefect-gitlab for more information. 证据：`src/integrations/prefect-gitlab/README.md`
- **prefect-kubernetes**（documentation）：See the docs at https://docs.prefect.io/integrations/prefect-kubernetes https://docs.prefect.io/integrations/prefect-kubernetes for more information. 证据：`src/integrations/prefect-kubernetes/README.md`
- **prefect-ray**（documentation）：See the docs at https://docs.prefect.io/integrations/prefect-ray https://docs.prefect.io/integrations/prefect-ray for more information. 证据：`src/integrations/prefect-ray/README.md`
- **prefect-redis**（documentation）：For more information about how to use Prefect, please refer to the Prefect documentation https://docs.prefect.io/ . 证据：`src/integrations/prefect-redis/README.md`
- **Integrate shell commands into your workflows with prefect-shell**（documentation）：Integrate shell commands into your workflows with prefect-shell 证据：`src/integrations/prefect-shell/README.md`
- **prefect-snowflake**（documentation）：See the docs at https://docs.prefect.io/integrations/prefect-snowflake https://docs.prefect.io/integrations/prefect-snowflake for more information. 证据：`src/integrations/prefect-snowflake/README.md`
- **prefect-sqlalchemy**（documentation）：See the docs at https://docs.prefect.io/integrations/prefect-sqlalchemy https://docs.prefect.io/integrations/prefect-alchemy for more information. 证据：`src/integrations/prefect-sqlalchemy/README.md`
- **Schema Forms**（documentation）：Schema Forms Prefect uses schema generated forms for several features: 证据：`ui-v2/src/components/schemas/readme.md`
- **Adding Icons**（documentation）：Import new icon from https://lucide.dev/icons/ in ./constants and add entry alphabetically in ICONS 证据：`ui-v2/src/components/ui/icons/README.md`
- **Readme**（documentation）：The icons in this folder are converted into textures and applied to sprites for use in the Pixi application. 证据：`ui-v2/src/graphs/textures/icons/README.md`
- **Package**（package_manifest）：{ "name": "prefect-ui", "private": true, "version": "0.0.0", "type": "module", "engines": { "node": " =22.13.0 <23" }, "scripts": { "analyze": "vite build && npx vite-bundle-visualizer", "build": "tsc -b && vite build", "dev": "vite", "test": "vitest", "test:e2e": "playwright test", "test:e2e:ui": "playwright test --ui", "test:e2e:debug": "playwright test --debug", "lint": "eslint .", "lint:fix": "eslint . --fix", "check": "biome check --write", "check:staged": "biome check --write --staged --no-errors-on-unmatched", "format": "biome format --write", "preview": "vite preview", "service-sync": "uv run ../scripts/generate oss openapi schema.py && npx openapi-typescript oss schema.json -o src/… 证据：`ui-v2/package.json`
- **Package**（package_manifest）：{ "name": "@prefecthq/ui", "version": "2.8.0", "private": true, "scripts": { "serve": "vite --host", "build": "vite build", "lint": "eslint src --ext .js,.ts,.vue", "lint:fix": "eslint src --fix", "validate:types": "vue-tsc --noEmit" }, "dependencies": { "@prefecthq/prefect-design": "2.16.4", "@prefecthq/prefect-ui-library": "3.15.0", "@prefecthq/vue-charts": "^2.1.1", "@prefecthq/vue-compositions": "^1.12.2", "@types/lodash.debounce": "4.0.9", "axios": "^1.16.0", "lodash.debounce": "4.0.8", "lodash.merge": "^4.6.2", "tailwindcss": "3.4.16", "vue": "3.5.24", "vue-router": "4.5.0" }, "devDependencies": { "@prefecthq/eslint-config": "1.0.32", "@types/lodash.merge": "^4.6.9", "@vitejs/plugin-v… 证据：`ui/package.json`
- **Package**（package_manifest）：{ "dependencies": { "aws-cdk": "2.1122.0" } } 证据：`src/integrations/prefect-aws/infra/package.json`
- **Guiding Principles**（documentation）：Prefect is a workflow orchestration platform that coordinates and observes data pipelines. It provides a Python SDK for building workflows, a server backend for orchestration, and a web-based UI for managing and monitoring workflows. 证据：`AGENTS.md`
- **prefect-client Package Build**（documentation）：Build configuration for prefect-client , a lightweight subset of prefect published as a separate PyPI package. This directory does not contain source code — it selects files from src/prefect/ and repackages them. 证据：`client/AGENTS.md`
- **Integration Tests**（documentation）：End-to-end tests that require a live Prefect server and verify the full orchestration pipeline across process boundaries. 证据：`integration-tests/AGENTS.md`
- **Prefect UI v2 - React Migration**（documentation）：This is the new React-based Prefect UI, migrating from the legacy Vue application with minimal visual and functional changes. 证据：`ui-v2/AGENTS.md`
- **Prefect Integrations**（documentation）：Official integrations extending Prefect with external services and platforms. Each integration is a separate PyPI package with its own version, dependencies, and test suite. 证据：`src/integrations/AGENTS.md`
- **prefect-aws**（documentation）：AWS integration for Prefect: blocks, tasks, and an ECS worker for running flows on Fargate. 证据：`src/integrations/prefect-aws/AGENTS.md`
- **prefect dbt.core**（documentation）：Core execution engine for the prefect-dbt integration. Contains the runner, orchestrator, manifest parser, hooks, cache, executor, and artifact logic. 证据：`src/integrations/prefect-dbt/prefect_dbt/core/AGENTS.md`
- **Core Prefect SDK**（documentation）：The foundation for building and executing workflows with Python. This covers the SDK layer — decorators, engines, state management, and client-side abstractions. 证据：`src/prefect/AGENTS.md`
- **Blocks**（documentation）：Reusable, server-persisted configuration objects for connecting workflows to external services. Blocks store credentials and settings securely and are shared across runs via the Prefect server. 证据：`src/prefect/blocks/AGENTS.md`
- **CLI**（documentation）：Prefect command-line interface, powered by cyclopts . 证据：`src/prefect/cli/AGENTS.md`
- **Client SDK**（documentation）：HTTP client for communicating with Prefect server and Prefect Cloud. 证据：`src/prefect/client/AGENTS.md`
- **Concurrency**（documentation）：Client-side concurrency limit enforcement: acquires and releases slots against server-defined limits, maintains leases during execution, and tracks events for traceability. 证据：`src/prefect/concurrency/AGENTS.md`
- **Deployments**（documentation）：YAML-driven configuration for packaging, publishing, and triggering flow runs from infrastructure. 证据：`src/prefect/deployments/AGENTS.md`
- **Events Client-Side**（documentation）：Client-side event system for emitting, subscribing to, and defining automations on Prefect events. 证据：`src/prefect/events/AGENTS.md`
- **Prefect Logging**（documentation）：Structured logging infrastructure for Prefect: handler pipeline, API log shipping, run-context-aware loggers, and WebSocket log streaming. 证据：`src/prefect/logging/AGENTS.md`
- **Runner**（documentation）：Thin facade over single-responsibility extracted classes. New behavior belongs in the extracted classes, not the Runner. 证据：`src/prefect/runner/AGENTS.md`
- **Prefect Server**（documentation）：Orchestration backend managing flow runs, scheduling, and state tracking. This is the source of truth for all state transitions. 证据：`src/prefect/server/AGENTS.md`
- **Settings**（documentation）：This directory contains Prefect's settings system built on Pydantic settings. 证据：`src/prefect/settings/AGENTS.md`
- **Utilities**（documentation）：General-purpose helpers and cross-cutting tools used throughout the Prefect SDK and server. 证据：`src/prefect/utilities/AGENTS.md`
- **asyncutils**（documentation）：Async/sync bridging, thread coordination, and concurrency primitives used throughout the SDK. 证据：`src/prefect/utilities/asyncutils/AGENTS.md`
- **callables**（documentation）：Function signature introspection, parameter coercion, and JSON Schema generation for flow and task parameters. 证据：`src/prefect/utilities/callables/AGENTS.md`
- **engine**（documentation）：Result-to-state linking, SIGTERM bridge management, and helpers shared by the flow/task engines. 证据：`src/prefect/utilities/engine/AGENTS.md`
- **filesystem**（documentation）：File path utilities, pathspec-based filtering, and local filesystem context managers. 证据：`src/prefect/utilities/filesystem/AGENTS.md`
- **processutils**（documentation）：Subprocess execution, output streaming, command serialization, and signal forwarding. 证据：`src/prefect/utilities/processutils/AGENTS.md`
- **schema tools**（documentation）：Hydration and validation of prefect kind template structures used in deployment parameters and automation action payloads. 证据：`src/prefect/utilities/schema_tools/AGENTS.md`
- **templating**（documentation）：Placeholder detection and value application for Prefect's {{ }} templating used in block references, workspace variables, and parameter templating. 证据：`src/prefect/utilities/templating/AGENTS.md`
- **Workers**（documentation）：Work-pool-based execution layer that polls for flow runs and submits them to infrastructure. 证据：`src/prefect/workers/AGENTS.md`
- **E2E Testing Playwright**（documentation）：This directory contains end-to-end tests using Playwright to test complete user workflows. 证据：`ui-v2/e2e/AGENTS.md`
- 其余 19 条证据见 `AI_CONTEXT_PACK.json` 或 `EVIDENCE_INDEX.json`。

## 宿主 AI 必须遵守的规则

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

- **项目概览**：importance `high`
  - source_paths: README.md, benches/README.md, client/README.md, client/pyproject.toml, load_testing/README.md
- **Src 模块**：importance `high`
  - source_paths: src/integrations/prefect-kubernetes/integration_tests/src/prefect_kubernetes_integration_tests/__init__.py, src/integrations/prefect-kubernetes/integration_tests/src/prefect_kubernetes_integration_tests/conftest.py, src/integrations/prefect-kubernetes/integration_tests/src/prefect_kubernetes_integration_tests/py.typed, src/integrations/prefect-kubernetes/integration_tests/src/prefect_kubernetes_integration_tests/test_crash_detection.py, src/integrations/prefect-kubernetes/integration_tests/src/prefect_kubernetes_integration_tests/test_events.py
- **Server 模块**：importance `high`
  - source_paths: src/prefect/settings/models/server/__init__.py, src/prefect/settings/models/server/api.py, src/prefect/settings/models/server/concurrency.py, src/prefect/settings/models/server/database.py, src/prefect/settings/models/server/deployments.py
- **Api 模块**：importance `high`
  - source_paths: src/prefect/server/api/__init__.py, src/prefect/server/api/_ui_static.py, src/prefect/server/api/admin.py, src/prefect/server/api/artifacts.py, src/prefect/server/api/automations.py

## Repo Inspection Evidence / 源码检查证据

- repo_clone_verified: true
- repo_inspection_verified: true
- repo_commit: `441d5e112d263706d963612aabc8f252df94a42e`
- inspected_files: `Dockerfile`, `README.md`, `pyproject.toml`, `uv.lock`, `docs/AGENTS.md`, `docs/conftest.py`, `docs/contribute/code-of-conduct.mdx`, `docs/contribute/contribute-integrations.mdx`, `docs/contribute/dev-contribute.mdx`, `docs/contribute/develop-a-new-worker-type.mdx`, `docs/contribute/docs-contribute.mdx`, `docs/contribute/index.mdx`, `docs/contribute/styles-practices.mdx`, `docs/docs.json`, `docs/integrations/index.md`, `docs/integrations/integrations.mdx`, `docs/integrations/prefect-aws/api-ref/prefect_aws-assume_role_parameters.mdx`, `docs/integrations/prefect-aws/api-ref/prefect_aws-batch.mdx`, `docs/integrations/prefect-aws/api-ref/prefect_aws-bundles-__init__.mdx`, `docs/integrations/prefect-aws/api-ref/prefect_aws-bundles-execute.mdx`

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: 失败模式：installation: 3.7.4 - From cradle to grave

- Trigger: Developers should check this installation risk before relying on the project: 3.7.4 - From cradle to grave
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: 3.7.4 - From cradle to grave. Context: Observed when using python
- Why it matters: Upgrade or migration may change expected behavior: 3.7.4 - From cradle to grave
- Evidence: failure_mode_cluster:github_release | https://github.com/PrefectHQ/prefect/releases/tag/3.7.4 | 3.7.4 - From cradle to grave
- 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: 失败模式：installation: 3.7.4.dev3: Nightly Development Release

- Trigger: Developers should check this installation risk before relying on the project: 3.7.4.dev3: Nightly Development Release
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: 3.7.4.dev3: Nightly Development Release. Context: Observed during installation or first-run setup.
- Why it matters: Upgrade or migration may change expected behavior: 3.7.4.dev3: Nightly Development Release
- Evidence: failure_mode_cluster:github_release | https://github.com/PrefectHQ/prefect/releases/tag/3.7.4.dev3 | 3.7.4.dev3: Nightly Development Release
- 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：Concurrent subflow calls adopt the WRONG subflow's persisted result on parent flow retry (positional dynamic_key race)

- 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：Concurrent subflow calls adopt the WRONG subflow's persisted result on parent flow retry (positional dynamic_key race)
- Host AI rule: 来源显示可能已有修复、规避或版本变化，说明书中必须标注适用版本。
- Why it matters: 可能增加新用户试用和生产接入成本。
- Evidence: community_evidence:github | https://github.com/PrefectHQ/prefect/issues/22259 | 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：Leaked `ephemeral-*` Redis consumer groups pin the `events` stream and prevent trimming after ungraceful shutdowns

- 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：Leaked `ephemeral-*` Redis consumer groups pin the `events` stream and prevent trimming after ungraceful shutdowns
- Host AI rule: 来源显示可能已有修复、规避或版本变化，说明书中必须标注适用版本。
- Why it matters: 可能阻塞安装或首次运行。
- Evidence: community_evidence:github | https://github.com/PrefectHQ/prefect/issues/22136 | 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 5: 失败模式：configuration: 3.7.4.dev4: Nightly Development Release

- Trigger: Developers should check this configuration risk before relying on the project: 3.7.4.dev4: Nightly Development Release
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: 3.7.4.dev4: Nightly Development Release. Context: Source discussion did not expose a precise runtime context.
- Why it matters: Upgrade or migration may change expected behavior: 3.7.4.dev4: Nightly Development Release
- Evidence: failure_mode_cluster:github_release | https://github.com/PrefectHQ/prefect/releases/tag/3.7.4.dev4 | 3.7.4.dev4: Nightly Development Release
- 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：configuration: 3.7.5 - No run left CANCELLING

- Trigger: Developers should check this configuration risk before relying on the project: 3.7.5 - No run left CANCELLING
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: 3.7.5 - No run left CANCELLING. Context: Observed when using python, docker
- Why it matters: Upgrade or migration may change expected behavior: 3.7.5 - No run left CANCELLING
- Evidence: failure_mode_cluster:github_release | https://github.com/PrefectHQ/prefect/releases/tag/3.7.5 | 3.7.5 - No run left CANCELLING
- 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: 3.7.5.dev1: Nightly Development Release

- Trigger: Developers should check this configuration risk before relying on the project: 3.7.5.dev1: Nightly Development Release
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: 3.7.5.dev1: Nightly Development Release. Context: Source discussion did not expose a precise runtime context.
- Why it matters: Upgrade or migration may change expected behavior: 3.7.5.dev1: Nightly Development Release
- Evidence: failure_mode_cluster:github_release | https://github.com/PrefectHQ/prefect/releases/tag/3.7.5.dev1 | 3.7.5.dev1: Nightly Development Release
- 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: 3.7.5.dev2: Nightly Development Release

- Trigger: Developers should check this configuration risk before relying on the project: 3.7.5.dev2: Nightly Development Release
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: 3.7.5.dev2: Nightly Development Release. Context: Source discussion did not expose a precise runtime context.
- Why it matters: Upgrade or migration may change expected behavior: 3.7.5.dev2: Nightly Development Release
- Evidence: failure_mode_cluster:github_release | https://github.com/PrefectHQ/prefect/releases/tag/3.7.5.dev2 | 3.7.5.dev2: Nightly Development Release
- 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: 3.7.5.dev3: Nightly Development Release

- Trigger: Developers should check this configuration risk before relying on the project: 3.7.5.dev3: Nightly Development Release
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: 3.7.5.dev3: Nightly Development Release. Context: Observed when using python, docker
- Why it matters: Upgrade or migration may change expected behavior: 3.7.5.dev3: Nightly Development Release
- Evidence: failure_mode_cluster:github_release | https://github.com/PrefectHQ/prefect/releases/tag/3.7.5.dev3 | 3.7.5.dev3: Nightly Development Release
- 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: 3.7.5.dev4: Nightly Development Release

- Trigger: Developers should check this configuration risk before relying on the project: 3.7.5.dev4: Nightly Development Release
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: 3.7.5.dev4: Nightly Development Release. Context: Observed when using python
- Why it matters: Upgrade or migration may change expected behavior: 3.7.5.dev4: Nightly Development Release
- Evidence: failure_mode_cluster:github_release | https://github.com/PrefectHQ/prefect/releases/tag/3.7.5.dev4 | 3.7.5.dev4: Nightly Development Release
- 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。