# wet-mcp - 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 wet-mcp 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：`skills/compare/SKILL.md`, `skills/fact-check/SKILL.md`, `skills/lock-project-stack/SKILL.md`, `skills/research-topic/SKILL.md` Claim：`clm_0004` 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：`skills/compare/SKILL.md`, `skills/fact-check/SKILL.md`, `skills/lock-project-stack/SKILL.md`, `skills/research-topic/SKILL.md` Claim：`clm_0001` supported 0.86
- **多宿主安装与分发**（需要安装后验证）：项目包含插件或 marketplace 配置，说明它面向一个或多个 AI 宿主的安装和分发。 证据：`.claude-plugin/plugin.json` Claim：`clm_0002` 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：`CONTRIBUTING.md` Claim：`clm_0003` supported 0.86

## 怎么开始

- `git clone https://github.com/YOUR_USERNAME/wet-mcp` 证据：`CONTRIBUTING.md` Claim：`clm_0005` supported 0.86

## 继续前判断卡

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

### 30 秒判断

- **现在怎么做**：需要管理员/安全审批
- **最小安全下一步**：先跑 Prompt Preview；若涉及凭证或企业环境，先审批再试装
- **先别相信**：研究结论、引用和实验结果不能在安装前相信。
- **继续会触碰**：研究判断、命令执行、宿主 AI 配置

### 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：`skills/compare/SKILL.md`, `skills/fact-check/SKILL.md`, `skills/lock-project-stack/SKILL.md`, `skills/research-topic/SKILL.md` Claim：`clm_0004` 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：`skills/compare/SKILL.md`, `skills/fact-check/SKILL.md`, `skills/lock-project-stack/SKILL.md`, `skills/research-topic/SKILL.md` Claim：`clm_0001` supported 0.86
- **能力存在：多宿主安装与分发**（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-plugin/plugin.json` Claim：`clm_0002` supported 0.86
- **能力存在：命令行启动或安装流程**（supported）：可以相信项目包含这类能力线索；是否适合你的具体任务仍要试用或安装后验证。 证据：`CONTRIBUTING.md` Claim：`clm_0003` 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：`CONTRIBUTING.md` Claim：`clm_0005` supported 0.86

### 现在还不能相信

- **研究结论、引用和实验结果不能在安装前相信。**（unverified）：研究 Skill 可以组织问题和路径，但不能替代真实资料检索、论文核验和实验复现。
- **是否适合你的具体研究领域不能直接相信。**（unverified）：Skill 覆盖很多研究主题，不代表对你的领域、资料要求和可信度标准足够。
- **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-plugin/plugin.json`, `AGENTS.md`, `CLAUDE.md`, `skills/compare/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。 证据：`.claude-plugin/plugin.json`
- **项目输出质量是否满足用户具体任务？**（unverified）：安装前预览只能展示流程和边界，不能替代真实评测。

### 继续会触碰什么

- **研究判断**：问题拆解、资料路径、实验路径、结论结构和可信度判断。 原因：研究型 Skill 可能让输出看起来更专业，但不能替代真实证据核验。
- **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：`CONTRIBUTING.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-plugin/plugin.json`, `AGENTS.md`, `CLAUDE.md`, `skills/compare/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：`.claude-plugin/plugin.json`, `CONTRIBUTING.md`
- **环境变量 / API Key**：项目入口文档明确出现 API key、token、secret 或账号凭证配置。 原因：如果真实安装需要凭证，应先使用测试凭证并经过权限/合规判断。 证据：`.claude-plugin/plugin.json`, `AGENTS.md`, `CHANGELOG.md`, `CLAUDE.md` 等
- **宿主 AI 上下文**：AI Context Pack、Prompt Preview、Skill 路由、风险规则和项目事实。 原因：导入上下文会影响宿主 AI 后续判断，必须避免把未验证项包装成事实。

### 最小安全下一步

- **先跑 Prompt Preview**：先验证它能否正确界定研究问题和证据边界，不要先相信研究输出。（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。）
- **不要使用真实生产凭证**：环境变量/API key 一旦进入宿主或工具链，可能产生账号和合规风险。（适用：出现 API、TOKEN、KEY、SECRET 等环境线索时。）
- **安装后只验证一个最小任务**：先验证加载、兼容、输出质量和回滚，再决定是否深用。（适用：准备从试用进入真实工作流时。）

### 退出方式

- **保留安装前状态**：记录原始宿主配置和项目状态，后续才能判断是否可恢复。
- **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。
- **保留资料和结论核验清单**：如果后续发现引用或实验路径不可靠，可以回到证据边界阶段重新校验。
- **记录安装命令和写入路径**：没有明确卸载说明时，至少要知道哪些目录或配置需要手动清理。
- **准备撤销测试 API key 或 token**：测试凭证泄露或误用时，可以快速止损。
- **如果没有回滚路径，不进入主力环境**：不可回滚是继续前阻断项，不应靠信任或运气继续。

## 哪些只能预览

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

## 哪些必须安装后验证

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

## 边界与风险判断卡

- **把安装前预览误认为真实运行**：用户可能高估项目已经完成的配置、权限和兼容性验证。 处理方式：明确区分 prompt_preview_can_do 与 runtime_required。 Claim：`clm_0006` inferred 0.45
- **宿主 AI 插件或 Skill 规则冲突**：新规则可能改变用户现有宿主 AI 的工作方式。 处理方式：安装前先检查插件 manifest 和 Skill 文件，必要时隔离测试。 证据：`.claude-plugin/plugin.json` Claim：`clm_0007` supported 0.86
- **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：`CONTRIBUTING.md` Claim：`clm_0008` 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：`skills/compare/SKILL.md`, `skills/fact-check/SKILL.md`, `skills/lock-project-stack/SKILL.md`, `skills/research-topic/SKILL.md` Claim：`clm_0001` supported 0.86
- **多宿主安装与分发**：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：`.claude-plugin/plugin.json` Claim：`clm_0002` 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：`CONTRIBUTING.md` Claim：`clm_0003` supported 0.86

### 上下文规模

- 文件总数：113
- 重要文件覆盖：40/113
- 证据索引条目：68
- 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 wet-mcp 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 wet-mcp 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 wet-mcp 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。

- **compare**（skill）：Structured comparison of 2+ alternatives with consistent criteria and decision matrix 激活提示：当用户任务与“compare”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：`skills/compare/SKILL.md`
- **fact-check**（skill）：Verify a claim using adversarial search — find both supporting AND contradicting evidence 激活提示：当用户任务与“fact-check”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：`skills/fact-check/SKILL.md`
- **lock-project-stack**（skill）：Detect a project's manifest pyproject.toml / package.json / go.mod / Cargo.toml , pin its library set into wet-mcp's Cabinets project context, then route subsequent docs queries to the locked versions automatically. 激活提示：当用户任务与“lock-project-stack”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：`skills/lock-project-stack/SKILL.md`
- **research-topic**（skill）：Multi-step research orchestration. Use when user asks "research X", "summarize current state of Y", "what's the latest on Z", or compares approaches. Calls extract action="agent" which searches the web, extracts top results, then synthesises a citation-preserving Markdown answer with one configured LLM. 激活提示：当用户任务与“research-topic”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：`skills/research-topic/SKILL.md`

## 证据索引

- 共索引 68 条证据。

- **AGENTS.md - wet-mcp**（documentation）：Python MCP Server: web search, content extraction, library docs, structured extraction. Xem AGENTS.md va README.md de hieu architecture va configuration. 证据：`AGENTS.md`
- **wet-mcp**（documentation）：Python MCP Server: web search, content extraction, library docs, structured extraction. Xem AGENTS.md va README.md de hieu architecture va configuration. 证据：`CLAUDE.md`
- **Package**（package_manifest）：{ "name": "wet-mcp-worker", "private": true, "type": "module", "scripts": { "test:worker": "vitest run tests/worker.test.ts", "cf:dryrun": "wrangler deploy --dry-run --outdir .wrangler-dryrun", "cf:deploy": "node scripts/cf-deploy.mjs" }, "devDependencies": { "@cloudflare/containers": "^0.3.7", "@cloudflare/workers-types": "^4.20260621.1", "typescript": "^6.0.3", "vitest": "^4.1.9", "wrangler": "^4.103.0" } } 证据：`package.json`
- **Contributing to wet-mcp**（documentation）：Thank you for your interest in contributing to wet-mcp! This guide will help you get started. 证据：`CONTRIBUTING.md`
- **Compare**（skill_instruction）：Structured comparison of 2+ alternatives. Enforces a comparison matrix with consistent criteria so every alternative is evaluated on the SAME data points. 证据：`skills/compare/SKILL.md`
- **Fact Check**（skill_instruction）：Verify a claim by actively searching for BOTH supporting and contradicting evidence. Counteracts LLM confirmation bias by enforcing adversarial search. 证据：`skills/fact-check/SKILL.md`
- **lock-project-stack**（skill_instruction）：Bind a project to its declared dependency versions so every search action="docs query" call from inside that project surfaces docs for the exact versions you pinned, instead of "latest". Mirrors the Cabinets workflow from spec section 4.3. 证据：`skills/lock-project-stack/SKILL.md`
- **research-topic**（skill_instruction）：Drive wet-mcp's extract action="agent" to answer a research question end to end: one search round + concurrent extracts of the top hits + a single LLM synthesis pass that preserves numbered N citations matching the returned sources. 证据：`skills/research-topic/SKILL.md`
- **Plugin**（structured_config）：{ "name": "wet-mcp", "description": "Web search, content extraction, and media download", "version": "3.3.0-beta.21", "userConfig": { "EMBEDDING MODELS": { "type": "string", "title": "Embedding model chain optional ", "description": "CSV 'provider/model,provider/model' order = litellm fallback ; provider inferred from prefix. Empty = local ONNX qwen3-embed . Example: jina ai/jina-embeddings-v5-text-small,gemini/gemini-embedding-001", "required": false }, "RERANK MODELS": { "type": "string", "title": "Rerank model chain optional ", "description": "CSV 'provider/model,...'; provider inferred from prefix. Empty = local ONNX cross-encoder. Example: jina ai/jina-reranker-v3", "required": false }… 证据：`.claude-plugin/plugin.json`
- **License**（source_file）：Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files the "Software" , to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: 证据：`LICENSE`
- **wet-mcp Architecture**（documentation）：v2.0.0: extract.agent + extract.interact shipped; media.analyze removed BREAKING . 证据：`docs/ARCHITECTURE.md`
- **wet-mcp Benchmarks**（documentation）：v2.0.0: extract.agent + extract.interact shipped BREAKING . 证据：`docs/BENCHMARKS.md`
- **Cloudflare worker template FROZEN for P3 replication**（documentation）：Cloudflare worker template FROZEN for P3 replication 证据：`docs/cf-template.md`
- **wet-mcp Docs Search Context7-parity**（documentation）：wet-mcp Docs Search Context7-parity 证据：`docs/docs-search.md`
- **extract action="interact" Guide**（documentation）：Drive a page interactively via patchright undetected Playwright fork with a small action language. Useful for surfaces that need a few targeted user actions before content becomes visible: 证据：`docs/interact.md`
- **Migration: wet v1.x.y - v2.0.0**（documentation）：This guide covers what changed in the v2.0.0 BREAKING release and how to upgrade existing wet-mcp deployments cleanly. 证据：`docs/migration.md`
- **Docs Sync Passport-Style Backend**（documentation）：wet-mcp ships a backend-pluggable sync layer that lets you carry the indexed docs cache docs.db across machines without re-crawling upstream documentation. Two backends are available: per-user Google Drive default uvx mode and operator-provisioned S3 / R2 / B2 / MinIO HTTP / Docker deploy mode . 证据：`docs/passport.md`
- **WET MCP Server - Config Tool**（documentation）：Server configuration and management. 证据：`src/wet_mcp/docs/config.md`
- **extract Tool Documentation**（documentation）：Extract content from web pages, crawl sites, or map site structure. 证据：`src/wet_mcp/docs/extract.md`
- **media Tool Documentation**（documentation）：Media discovery and download from web pages. 证据：`src/wet_mcp/docs/media.md`
- **search Tool Documentation**（documentation）：Search the web, academic papers, or library documentation. 证据：`src/wet_mcp/docs/search.md`
- **Docs 002 Libraries**（source_file）：revision = "docs 002 libraries" down revision = "docs 001 baseline" branch labels = None depends on = None logger = logging.getLogger "alembic.runtime.migration" def existing columns table: str - set str ⋮---- bind = op.get bind rows = bind.exec driver sql ⋮---- def existing indexes table: str - set str def add column if missing table: str, name: str, ddl: str - None def upgrade - None def downgrade - None 证据：`src/wet_mcp/alembic/versions/docs_002_libraries.py`
- **Docs 004 Chunk Summaries**（source_file）：revision = "docs 004 chunk summaries" down revision = "docs 003 project context" branch labels = None depends on = None def upgrade - None ⋮---- existing cols = { ⋮---- def downgrade - None 证据：`src/wet_mcp/alembic/versions/docs_004_chunk_summaries.py`
- **Credential State**（source_file）：SERVER NAME = "wet-mcp" PLUGIN NAME = "wet" current sub: contextvars.ContextVar str None = contextvars.ContextVar SPAWN CLEANUP S = 5.0 CLOUD KEYS = class CredentialState Enum ⋮---- AWAITING SETUP = "awaiting setup" SETUP IN PROGRESS = "setup in progress" CONFIGURED = "configured" LOCAL = "local" state = CredentialState.AWAITING SETUP setup url: str None = None active handle: Any None = None on gdrive complete: Callable , None None = None on gdrive failed: Callable str, str , None None = None def set gdrive complete callback cb: Callable , None - None ⋮---- on gdrive complete = cb ⋮---- def set gdrive failed callback cb: Callable str, str , None - None ⋮---- on gdrive failed = cb ⋮---- def… 证据：`src/wet_mcp/credential_state.py`
- **bs4 missing — regex fallback good enough for well-formed pages .**（source_file）：CODE BLOCK RE = re.compile r" ", re.DOTALL ⋮---- bs4 missing — regex fallback good enough for well-formed pages . 证据：`src/wet_mcp/sources/_smart_chunks.py`
- **smart chunks output uses markdown or clean text ; document**（source_file）：PROVIDER KEYS = LLM PROVIDER KEYS DEFAULT MAX URLS = 5 HARD MAX URLS = 20 DEFAULT TOKEN BUDGET = 10000 CHARS PER TOKEN = 4 EXTRACT CONCURRENCY = 3 def detect llm provider - str None def no provider error - str def clamp max urls max urls: int - int def truncate for budget text: str, char budget: int - str ⋮---- per extract chars = max citation blocks: list str = ⋮---- url = extract.get "url", "" smart chunks output uses markdown or clean text ; document bridge uses content . Try them in priority order. body = body = truncate for budget body, per extract chars title = extract.get "metadata" or {} .get "title" or extract.get ⋮---- sources block = "\n\n".join citation blocks ⋮---- async def ex… 证据：`src/wet_mcp/sources/agent_orchestrator.py`
- **ReadTheDocs**（source_file）：DISCOVERY VERSION = 27 def github headers - dict str, str ⋮---- headers: dict str, str = {} token = os.environ.get "GITHUB TOKEN" or os.environ.get "GH TOKEN" ⋮---- def apply version to url url: str, version: str None - str ⋮---- """Apply version to docs URL if applicable. Handles ReadTheDocs /en/latest/ - /en/{version}/ and docs.rs /latest/ - /{version}/ patterns. """ ⋮---- ReadTheDocs ⋮---- url = re.sub r"/en/ latest stable /", f"/en/{version}/", url docs.rs ⋮---- url = re.sub r"/latest/", f"/{version}/", url ⋮---- --------------------------------------------------------------------------- Registry discovery — find docs URL from library name ⋮---- async def discover from npm name: str - d… 证据：`src/wet_mcp/sources/docs.py`
- **Search Backends**（source_file）：class SearchBackend Protocol class SearchHTTPError Exception ⋮---- def init self, provider: str, status code: int - None ⋮---- """Rotate keys for one search provider, advancing only on a key-specific failure 429/401/403 . Returns the tool-contract error envelope when keys are exhausted or a non-rotatable error occurs — type name only, never the exception text an httpx error may carry the request body, i.e. the key .""" ⋮---- except Exception as exc: tool contract: error string, never raise; type name only ⋮---- class SearxngBackend ⋮---- def init self, url: str - None ⋮---- class TavilyBackend ⋮---- URL = "https://api.tavily.com/search" def init self, keys: list str - None async def search… 证据：`src/wet_mcp/sources/search_backends.py`
- **Step 1: Extract source content**（source_file）：async def expand query query: str - list str ⋮---- mode = settings.resolve provider mode ⋮---- config = get llm config response = await acompletion text = response.choices 0 .message.content or "" alt queries = ⋮---- """Find pages similar to the given URL. Pipeline: 1. Extract source page content 2. Extract keywords via LLM or title fallback 3. Search SearXNG with keywords, excluding source domain 4. Return results as JSON string """ Step 1: Extract source content raw = await raw extract urls= url , format="markdown" pages = json.loads raw ⋮---- content = pages 0 .get "content", pages 0 .get "markdown", "" :3000 title = pages 0 .get "title", "" source domain = urlparse url .netloc Step 2: E… 证据：`src/wet_mcp/sources/search_strategies.py`
- **Attempt 2: json object mode simpler, wider provider support**（source_file）：MAX CONTENT CHARS = 50 000 SYSTEM PROMPT = ⋮---- llm kwargs = { ⋮---- response = await acompletion content = response.choices 0 .message.content ⋮---- Attempt 2: json object mode simpler, wider provider support ⋮---- llm kwargs, type: ignore invalid-argument-type ty: ignore invalid-argument-type dict unpacking ⋮---- mode = settings.resolve provider mode ⋮---- raw json = await raw extract urls, stealth=stealth pages = json.loads raw json ⋮---- Step 3: Combine content, truncate combined parts: list str = ⋮---- crawler.extract returns smart chunks keys clean text / markdown ; keep "content" as a backward-compat fallback. content = ⋮---- title = page.get "title", "" url = page.get "url", "" hea… 证据：`src/wet_mcp/sources/structured.py`
- **CHANGELOG**（documentation）：- Force real litellm to win unclecode-litellm file collision to restore catalog/LLM 1413 https://github.com/n24q02m/wet-mcp/pull/1413 , ddca53d https://github.com/n24q02m/wet-mcp/commit/ddca53dab0c959c298018f7ac3837d53f61746e1 证据：`CHANGELOG.md`
- **Contributor Covenant Code of Conduct**（documentation）：Contributor Covenant Code of Conduct 证据：`CODE_OF_CONDUCT.md`
- **Security Policy**（documentation）：Version Supported ------- ------------------ 3.x.x :white check mark: < 3.0 :x: 证据：`SECURITY.md`
- **.Infisical**（structured_config）：{ "workspaceId": "531b3027-70ca-4761-b149-9ec8fea80d4f", "defaultEnvironment": "", "gitBranchToEnvironmentMapping": null } 证据：`.infisical.json`
- **Glama**（structured_config）：{ "$schema": "https://glama.ai/mcp/schemas/server.json", "maintainers": "n24q02m" } 证据：`glama.json`
- **Renovate**（structured_config）：{ "$schema": "https://docs.renovatebot.com/renovate-schema.json", "extends": "config:recommended", ":semanticCommits" , "schedule": "before 5am" , "timezone": "Asia/Ho Chi Minh", "prConcurrentLimit": 5, "prHourlyLimit": 3, "labels": "dependencies" , "rangeStrategy": "bump", "lockFileMaintenance": { "enabled": true, "schedule": "before 5am" }, "packageRules": { "description": "Auto-merge non-major devDependencies", "matchDepTypes": "devDependencies" , "matchUpdateTypes": "minor", "patch" , "automerge": true, "automergeType": "pr" }, { "description": "Auto-merge patch production dependencies", "matchDepTypes": "dependencies" , "matchUpdateTypes": "patch" , "automerge": true, "automergeType":… 证据：`renovate.json`
- **Server**（structured_config）：{ "$schema": "https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json", "name": "io.github.n24q02m/wet-mcp", "description": "MCP server for web search, content extraction, academic research, and library docs.", "repository": { "url": "https://github.com/n24q02m/wet-mcp.git", "source": "github" }, "version": "3.3.0-beta.21", "packages": { "registryType": "pypi", "identifier": "wet-mcp", "version": "3.3.0-beta.21", "runtimeHint": "uvx", "runtimeArgs": "--python", "3.13" , "transport": { "type": "stdio" }, "environmentVariables": { "description": "Provider API keys format: PROVIDER API KEY:key,... . Select models per task with EMBEDDING MODELS / RERANK MODELS / LLM MODELS… 证据：`server.json`
- **Tsconfig**（structured_config）：{ "compilerOptions": { "target": "es2022", "module": "es2022", "moduleResolution": "bundler", "lib": "es2022" , "types": "@cloudflare/workers-types" , "strict": true, "skipLibCheck": true, "noEmit": true }, "include": "src/ / .ts", "tests/ / .ts" } 证据：`tsconfig.json`
- **V1 Phase2 Baseline**（structured_config）：{ "captured at": "2026-05-10", "git head": "22a93a05d0834e108886aa740dfdfa893890995d", "tag": "v2.31.0-beta.1", "phase": "phase-2-stable-candidate", "purpose": "Phase 3 regression-tolerance baseline spec section 3, +/- 10% .", "metrics": { "test suite": { "passed": 1677, "skipped": 34, "deselected": 141, "duration seconds": 65.93 }, "coverage": { "total pct": 92.86, "fail under gate": 92, "long term target pct": 95 }, "search p95 ms": null, "extract tier1 success rate": null, "docs query p95 ms": null, "docs query recall at 10": null, "startup time ms": null, "rss steady mb": null }, "phase 3 targets": { "extract agent p95 seconds for 5 url synthesis": 30, "extract interact per action p95 s… 证据：`docs/benchmarks/v1_phase2_baseline.json`
- **Tier1 Libraries**（structured_config）：{ " metadata": { "version": 1, "description": "Curated Tier 1 library index for wet-mcp Phase 2. Each entry seeds metadata only; doc chunks are ingested lazily on first docs query call or eagerly via scripts/build tier1 index.py ." }, "libraries": {"id": "react", "canonical name": "React", "tier": 1, "homepage": "https://react.dev", "github url": "https://github.com/facebook/react", "package managers": "npm", "yarn", "pnpm" }, {"id": "next", "canonical name": "Next.js", "tier": 1, "homepage": "https://nextjs.org", "github url": "https://github.com/vercel/next.js", "package managers": "npm" }, {"id": "vue", "canonical name": "Vue.js", "tier": 1, "homepage": "https://vuejs.org", "github url":… 证据：`src/wet_mcp/data/tier1_libraries.json`
- **.Coderabbit**（source_file）：language: "en-US" reviews: auto review: enabled: true drafts: false ignore usernames: - "n24q02m" - "dependabot bot " - "renovate bot " - "github-actions bot " - "devin-ai-integration bot " - "google-labs-jules bot " profile: "chill" request changes workflow: false high level summary: true auto incremental review: true issue enrichment: auto enrich: enabled: true planning: enabled: true chat: auto reply: false 证据：`.coderabbit.yaml`
- **Git**（source_file）：Cache .ruff cache .pytest cache pycache 证据：`.dockerignore`
- **Editor configuration**（source_file）：Editor configuration https://editorconfig.org 证据：`.editorconfig`
- **Auto detect text files and perform LF normalization**（source_file）：Auto detect text files and perform LF normalization text=auto 证据：`.gitattributes`
- **Python**（source_file）：Python pycache / .py cod $py.class .so .Python build/ develop-eggs/ dist/ downloads/ eggs/ .eggs/ lib/ lib64/ parts/ sdist/ var/ wheels/ .egg-info/ .installed.cfg .egg 证据：`.gitignore`
- **Mise configuration for wet-mcp**（source_file）：Mise configuration for wet-mcp https://mise.jdx.dev/ Setup: mise run setup 证据：`.mise.toml`
- **.Pre Commit Config**（source_file）：default language version: python: python3.13 repos: - repo: https://github.com/gitleaks/gitleaks rev: 83d9cd684c87d95d656c1458ef04895a7f1cbd8e hooks: - id: gitleaks name: gitleaks detect secrets - repo: https://github.com/astral-sh/ruff-pre-commit rev: e05c5c0818279e5ac248ac9e954431ba58865e61 hooks: - id: ruff name: ruff lint args: --fix, --target-version=py313 - id: ruff-format name: ruff format - repo: local hooks: - id: ty name: ty type check entry: uv run --no-sync ty check language: system types: python pass filenames: false - repo: local hooks: - id: pytest name: pytest Python entry: uv run --no-sync pytest --tb=short -q --timeout=30 language: system types: python pass filenames: fals… 证据：`.pre-commit-config.yaml`
- **Sync Claude Agents**（source_file）：set -e CLAUDE FILE="CLAUDE.md" AGENTS FILE="AGENTS.md" if ! -f "$CLAUDE FILE" ! -f "$AGENTS FILE" ; then echo "sync-claude-agents: missing $CLAUDE FILE or $AGENTS FILE in repo root." exit 1 fi CLAUDE HEAD=$ head -n 1 "$CLAUDE FILE" AGENTS HEAD=$ head -n 1 "$AGENTS FILE" if "$CLAUDE HEAD" != " wet-mcp" ; then echo "sync-claude-agents: CLAUDE.md line 1 must be ' wet-mcp' got: '$CLAUDE HEAD' ." exit 1 fi if "$AGENTS HEAD" != " AGENTS.md - wet-mcp" ; then echo "sync-claude-agents: AGENTS.md line 1 must be ' AGENTS.md - wet-mcp' got: '$AGENTS HEAD' ." exit 1 fi CLAUDE BODY=$ tail -n +2 "$CLAUDE FILE" AGENTS BODY=$ tail -n +2 "$AGENTS FILE" if "$CLAUDE BODY" != "$AGENTS BODY" ; then echo "sync-cl… 证据：`.pre-commit-hooks/sync-claude-agents.sh`
- **.python-version**（source_file）：3.13 证据：`.python-version`
- **syntax=docker/dockerfile:1**（source_file）：syntax=docker/dockerfile:1 Multi-stage build for wet-mcp Python 3.13 + SearXNG + Playwright chromium All-in-one: no external Docker or services needed 证据：`Dockerfile`
- **Docker Compose.Cloudflare**（source_file）：services: wet-mcp: build: . environment: - MCP TRANSPORT=http - MCP STORAGE BACKEND=cf-kv - MCP KV BASE URL=http://host.docker.internal:8787/kv - DOCS DB BACKEND=cf-d1 - MCP D1 BASE URL=http://host.docker.internal:8787/d1.internal - MCP VECTORIZE BASE URL=http://host.docker.internal:8787/vectorize.internal - MCP VECTORIZE IDX=wet-docs-vectors - EMBEDDING MODELS=jina ai/jina-embeddings-v5-text-small - RERANK MODELS=jina ai/jina-reranker-v3 - SEARCH BACKEND=tavily - PUBLIC URL=http://localhost:8787 - CREDENTIAL SECRET=${CREDENTIAL SECRET} - JINA AI API KEY=${JINA AI API KEY} - TAVILY API KEY=${TAVILY API KEY} ports: - "8085:8080" extra hosts: - "host.docker.internal:host-gateway" 证据：`docker-compose.cloudflare.yml`
- **Docker Compose.Http**（source_file）：services: wet-mcp: environment: - TRANSPORT MODE=http - CREDENTIAL SECRET=${CREDENTIAL SECRET} - MCP RELAY PASSWORD=${MCP RELAY PASSWORD} ports: - "8085:8080" volumes: - wet-data:/data volumes: wet-data: 证据：`docker-compose.http.yml`
- **Docker Compose**（source_file）：services: wet-mcp: image: n24q02m/wet-mcp:latest environment: - CACHE DIR=/data - DOWNLOAD DIR=/data/downloads volumes: - wet-data:/data volumes: wet-data: 证据：`docker-compose.yml`
- **Sessionstart Check Docs Freshness**（source_file）：set -euo pipefail DB PATH="${WET DOCS DB PATH:-$HOME/.wet-mcp/docs.db}" if ! -f "$DB PATH" ; then exit 0 fi if ! command -v sqlite3 /dev/null 2 &1; then exit 0 fi NOW=$ date +%s SEVEN DAYS AGO=$ NOW - 7 24 60 60 STALE=$ sqlite3 "$DB PATH" \ "SELECT COUNT FROM libraries WHERE tier = 1 AND last indexed at IS NULL OR last indexed at /dev/null echo 0 if "$STALE" -gt 0 ; then &2 echo "wet-mcp: $STALE Tier 1 libraries are 7 days stale." &2 echo "Refresh suggestion: config action='docs reindex', scope='tier1' " &2 echo " or run: uv run python scripts/build tier1 index.py" fi exit 0 证据：`hooks/SessionStart_check_docs_freshness.sh`
- **Userpromptsubmit Prewarm Docs Cache**（source_file）：set -euo pipefail PROMPT="${CLAUDE USER PROMPT:-}" if -z "$PROMPT" ; then if -t 0 ; then exit 0 fi PROMPT=$ cat true fi if -z "$PROMPT" ; then exit 0 fi TIER1 LIBRARIES= "fastapi" "pydantic" "starlette" "uvicorn" "react" "next" "vue" "svelte" "django" "flask" "numpy" "pandas" "polars" "scikit-learn" "pytorch" "tensorflow" "transformers" "langchain" "anthropic" "openai" CANDIDATES=$ { printf "%s\n" "$PROMPT" grep -oE 'from :space: + a-zA-Z0-9 .- +' awk '{print $2}' printf "%s\n" "$PROMPT" grep -oE 'import :space: + a-zA-Z0-9 .- +' awk '{print $2}' printf "%s\n" "$PROMPT" grep -oE 'require\ " a-zA-Z0-9 .@/- +"\ ' sed -E 's/require\ " ^" + "\ /\1/' printf "%s\n" "$PROMPT" grep -oE 'from :space… 证据：`hooks/UserPromptSubmit_prewarm_docs_cache.sh`
- **0001 Init Wet**（source_file）：CREATE TABLE IF NOT EXISTS libraries id TEXT PRIMARY KEY, name TEXT NOT NULL, docs url TEXT, registry TEXT, description TEXT, canonical name TEXT, homepage TEXT, github url TEXT, package managers TEXT, tier INTEGER NOT NULL DEFAULT 2, last indexed at REAL, total versions INTEGER NOT NULL DEFAULT 0, discovery version INTEGER DEFAULT 0, created at REAL NOT NULL, updated at REAL NOT NULL ; CREATE INDEX IF NOT EXISTS idx libraries name ON libraries name ; CREATE TABLE IF NOT EXISTS versions id TEXT PRIMARY KEY, library id TEXT NOT NULL, version TEXT NOT NULL DEFAULT 'latest', docs url TEXT, indexed at REAL, page count INTEGER DEFAULT 0, chunk count INTEGER DEFAULT 0, status TEXT DEFAULT 'pendin… 证据：`migrations/0001_init_wet.sql`
- **Shared web infrastructure SSRF, SearXNG, URL utils, ScrapingAgent + Crawl4AI transitive .**（source_file）：project name = "wet-mcp" version = "3.3.0-beta.21" description = "Open-source MCP Server for web search, extract, crawl, academic research, and library docs with embedded SearXNG" readme = "README.md" license = { text = "MIT" } authors = { name = "n24q02m", email = "quangminh2422004@gmail.com" } keywords = "mcp", "mcp-server", "model-context-protocol", "web-scraping", "searxng", "crawl4ai", "library-docs", "claude", "cursor", "copilot", "antigravity", "codex", "opencode" classifiers = "Development Status :: 4 - Beta", "Environment :: Console", "Intended Audience :: Developers", "License :: OSI Approved :: MIT License", "Operating System :: OS Independent", "Programming Language :: Python ::… 证据：`pyproject.toml`
- **Best-effort: only used by CI. Local users can ignore.**（source_file）：async def amain - int ⋮---- parser = argparse.ArgumentParser description="Eager Tier 1 docs ingestion" ⋮---- args = parser.parse args ⋮---- db = DocsDB args.db path, embedding dims=0 ⋮---- payload = load tier1 payload libraries = payload.get "libraries", ⋮---- metrics = started = time.time ⋮---- name = entry "id" t0 = time.time ⋮---- result = await ingest tier2 db, name duration = time.time - t0 ⋮---- except Exception as exc: pragma: no cover - script-level guard ⋮---- summary = { out path = args.db path.parent / "tier1 index metrics.json" ⋮---- Best-effort: only used by CI. Local users can ignore. ⋮---- def main - int 证据：`scripts/build_tier1_index.py`
- **Cf Deploy**（source_file）：/ Live Cloudflare Worker + Container deploy config-only by default . Companion to the "cf:dryrun" package.json script. Where cf:dryrun validates the bundle with wrangler deploy --dry-run , this ships it for real with wrangler deploy , applying the committed cost config instance type, sleepAfter, max instances to the live Worker + Container without rebuilding the image. Config resolution real CF resource IDs are never committed : 1. wrangler.deploy.jsonc - GITIGNORED deploy-time copy carrying the REAL account/KV/D1/Vectorize IDs. Preferred when present this is the live deploy path; the committed wrangler.jsonc keeps placeholder resource IDs . 2. wrangler.jsonc - committed config. Used as a f… 证据：`scripts/cf-deploy.mjs`
- **A distinctive query so the search path is exercised against the live web backend.**（source_file）：DEFAULT ENDPOINT = os.environ.get "CF ENDPOINT", "" A distinctive query so the search path is exercised against the live web backend. SEARCH QUERY = "cloudflare workers durable objects" def password - str ⋮---- pw = os.environ.get "RELAY PW" or os.environ.get "MCP RELAY PASSWORD" ⋮---- def creds - dict str, str ⋮---- creds: dict str, str = {} ⋮---- v = os.environ.get env name ⋮---- class SaveRetry Exception def get token endpoint: str, creds: dict str, str , , save retries: int = 8 - str ⋮---- last: Exception None = None ⋮---- last = e ⋮---- def get token once httpx, endpoint: str, creds: dict str, str - str ⋮---- verifier = base64.urlsafe b64encode secrets.token bytes 32 .rstrip b"=" .deco… 证据：`scripts/cf_full_flow.py`
- 其余 8 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：`AGENTS.md`, `CLAUDE.md`, `package.json`
- **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：`AGENTS.md`, `CLAUDE.md`, `package.json`

## 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。

- **项目概览与系统架构**：importance `high`
  - source_paths: README.md, docs/ARCHITECTURE.md, src/wet_mcp/server.py, src/wet_mcp/__main__.py, pyproject.toml
- **核心工具：搜索、抽取、媒体与配置**：importance `high`
  - source_paths: src/wet_mcp/sources/docs.py, src/wet_mcp/sources/crawler.py, src/wet_mcp/sources/_smart_chunks.py, src/wet_mcp/sources/_search_polish.py, src/wet_mcp/sources/search_backends.py
- **模型链、环境变量与 Cloudflare 部署**：importance `high`
  - source_paths: src/wet_mcp/config.py, src/wet_mcp/llm.py, src/wet_mcp/embedder.py, src/wet_mcp/reranker.py, src/wet_mcp/searxng_runner.py
- **迁移指南、运维实践与版本演进**：importance `high`
  - source_paths: docs/migration.md, docs/passport.md, docs/docs-search.md, docs/interact.md, docs/BENCHMARKS.md

## Repo Inspection Evidence / 源码检查证据

- repo_clone_verified: true
- repo_inspection_verified: true
- repo_commit: `a68e8573a831f64e288d3b0146908882c31e959b`
- inspected_files: `Dockerfile`, `README.md`, `docker-compose.yml`, `package.json`, `pyproject.toml`, `uv.lock`, `docs/ARCHITECTURE.md`, `docs/BENCHMARKS.md`, `docs/benchmarks/v1_phase2_baseline.json`, `docs/cf-template.md`, `docs/docs-search.md`, `docs/interact.md`, `docs/migration.md`, `docs/passport.md`, `src/wet_mcp/__init__.py`, `src/wet_mcp/__main__.py`, `src/wet_mcp/alembic/env.py`, `src/wet_mcp/alembic/versions/docs_001_baseline.py`, `src/wet_mcp/alembic/versions/docs_002_libraries.py`, `src/wet_mcp/alembic/versions/docs_003_project_context.py`

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

## Doramagic Pitfall Constraints / 踩坑约束

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

### Constraint 1: 需要 API Key 或环境变量

- Trigger: 项目说明中出现 API Key / 环境变量相关需求。
- Host AI rule: 提取必需 env 列表，确认是否有最小无密钥试用路径。
- Why it matters: 用户必须准备账号、额度或密钥；密钥配置错误会导致运行失败或泄漏风险。
- Evidence: packet_text.keyword_scan | https://github.com/n24q02m/wet-mcp | matched api key / env var keyword
- 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: Dependency Dashboard

- Trigger: Developers should check this installation risk before relying on the project: Dependency Dashboard
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: Dependency Dashboard. Context: Observed when using python, docker
- Why it matters: Developers may fail before the first successful local run: Dependency Dashboard
- Evidence: failure_mode_cluster:github_issue | https://github.com/n24q02m/wet-mcp/issues/231 | Dependency Dashboard
- 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: 失败模式：installation: v3.3.0-beta.18

- Trigger: Developers should check this installation risk before relying on the project: v3.3.0-beta.18
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: v3.3.0-beta.18. Context: Observed when using python, docker
- Why it matters: Upgrade or migration may change expected behavior: v3.3.0-beta.18
- Evidence: failure_mode_cluster:github_release | https://github.com/n24q02m/wet-mcp/releases/tag/v3.3.0-beta.18 | v3.3.0-beta.18
- 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: 可能修改宿主 AI 配置

- Trigger: 项目面向 Claude/Cursor/Codex/Gemini/OpenCode 等宿主，或安装命令涉及用户配置目录。
- Host AI rule: 列出会写入的配置文件、目录和卸载/回滚步骤。
- Why it matters: 安装可能改变本机 AI 工具行为，用户需要知道写入位置和回滚方法。
- Evidence: capability.host_targets | https://github.com/n24q02m/wet-mcp | host_targets=mcp_host, claude_code, claude, cursor
- 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: v3.3.0-beta.12

- Trigger: Developers should check this configuration risk before relying on the project: v3.3.0-beta.12
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: v3.3.0-beta.12. Context: Observed when using docker
- Why it matters: Upgrade or migration may change expected behavior: v3.3.0-beta.12
- Evidence: failure_mode_cluster:github_release | https://github.com/n24q02m/wet-mcp/releases/tag/v3.3.0-beta.12 | v3.3.0-beta.12
- 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: v3.3.0-beta.13

- Trigger: Developers should check this configuration risk before relying on the project: v3.3.0-beta.13
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: v3.3.0-beta.13. Context: Source discussion did not expose a precise runtime context.
- Why it matters: Upgrade or migration may change expected behavior: v3.3.0-beta.13
- Evidence: failure_mode_cluster:github_release | https://github.com/n24q02m/wet-mcp/releases/tag/v3.3.0-beta.13 | v3.3.0-beta.13
- 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: v3.3.0-beta.15

- Trigger: Developers should check this configuration risk before relying on the project: v3.3.0-beta.15
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: v3.3.0-beta.15. Context: Source discussion did not expose a precise runtime context.
- Why it matters: Upgrade or migration may change expected behavior: v3.3.0-beta.15
- Evidence: failure_mode_cluster:github_release | https://github.com/n24q02m/wet-mcp/releases/tag/v3.3.0-beta.15 | v3.3.0-beta.15
- 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: v3.3.0-beta.16

- Trigger: Developers should check this configuration risk before relying on the project: v3.3.0-beta.16
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: v3.3.0-beta.16. Context: Source discussion did not expose a precise runtime context.
- Why it matters: Upgrade or migration may change expected behavior: v3.3.0-beta.16
- Evidence: failure_mode_cluster:github_release | https://github.com/n24q02m/wet-mcp/releases/tag/v3.3.0-beta.16 | v3.3.0-beta.16
- 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: v3.3.0-beta.20

- Trigger: Developers should check this configuration risk before relying on the project: v3.3.0-beta.20
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: v3.3.0-beta.20. Context: Source discussion did not expose a precise runtime context.
- Why it matters: Upgrade or migration may change expected behavior: v3.3.0-beta.20
- Evidence: failure_mode_cluster:github_release | https://github.com/n24q02m/wet-mcp/releases/tag/v3.3.0-beta.20 | v3.3.0-beta.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: 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 | https://github.com/n24q02m/wet-mcp | 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。