# meridian-orbital - 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 meridian-orbital 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_0011` 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：`example-skills/hello-world/SKILL.md`, `example-skills/http-best-practices/SKILL.md`, `example-skills/markdown-formatting/SKILL.md`, `skills/agent-loop/SKILL.md` 等 Claim：`clm_0012` supported 0.86

## 它能做什么

- **Orbital Task Routing (Meridian MCP)**（可做安装前预览）：Domain-agnostic LLM-driven task routing that classifies candidate entities (tools, prompts, documents) using a 9-scalar orbital mechanics-inspired physics signature, scoring them against six celestial body classes (planet/moon/trojan/asteroid/comet/irregular) via deterministic argmax rules. 证据：`README.md`, `README.md`, `package.json` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86
- **Dual-Transport MCP Server**（需要安装后验证）：Self-contained MCP server supporting both stdio transport (for Claude Code, Cursor, Windsurf, Goose, Continue) and Streamable HTTP transport (for Grok, ChatGPT, Claude.ai connectors). Includes OAuth 2.1 + PKCE for hosted deployments. 证据：`README.md`, `README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86
- **Finance MCP (Binance/ Mercado Pago/ Coinbase)**（需要安装后验证）：Personal finance MCP on Cloudflare Workers with passkey-bootstrap OAuth. Manages Binance spot wallet, USDC-ARS conversions, USDC withdrawals to Coinbase, ARS withdrawals to Mercado Pago CVU. All writes use prepare-then-confirm pattern with daily spending caps. 证据：`finance-mcp/README.md`, `finance-mcp/README.md`, `finance-mcp/README.md`, `finance-mcp/package.json` Claim：`clm_0003` supported 0.86
- **Pharmacy MCP (Farmacias del Pueblo VTEX)**（需要安装后验证）：Anonymous-cart pharmacy MCP circumventing VTEX reCAPTCHA hostname enforcement. Provides catalog search, cart management, and browser hand-off URL for checkout. Live at botica.ask-meridian.uk. 证据：`pharmacy-mcp/README.md`, `pharmacy-mcp/README.md`, `pharmacy-mcp/package.json` Claim：`clm_0004` supported 0.86
- **Binance Proxy (Bright Data Tunnel)**（需要安装后验证）：HTTP proxy on Fly.io bridging CF Workers to Binance API via Bright Data CONNECT tunnel, solving geo-blocking issue. Static residential exit IP whitelisted on Binance API key. 证据：`binance-proxy/README.md`, `binance-proxy/README.md`, `binance-proxy/package.json` Claim：`clm_0005` supported 0.86
- **Helio Mirror (Solar Irradiance Triangulator)**（可做安装前预览）：6-stage GitHub Actions pipeline cross-correlating PSP solar-event signatures with JWST reflectance excursions on solar-system bodies to forecast 24h irradiance delivered to each body. 证据：`helio-mirror/README.md`, `helio-mirror/README.md` Claim：`clm_0006` supported 0.86
- **Photon Route (Continuous-Variable Photonic Retrieval)**（需要安装后验证）：Document/query retrieval using Strawberry Fields continuous-variable photonic encoding. Documents represented as Gaussian quantum states in Fock space; retrieval via Gaussian-state fidelity (Banchi-Braunstein-Pirandola). CV sister to qrouter (discrete-variable). 证据：`photon-route/README.md`, `photon-route/README.md` Claim：`clm_0007` supported 0.86
- **Design System / UI Kits**（可做安装前预览）：Canonical visual language: CSS tokens, type scale, motion primitives, celestial color system. Per-surface UI kits for landing, miniapp, docs, helio, helix, lens, photon-router surfaces. Tokens mirrored in live runtime CSS across all surfaces. 证据：`design-system/README.md`, `design-system/README.md`, `design-system/ui_kits/docs/README.md`, `design-system/ui_kits/miniapp/README.md` Claim：`clm_0008` supported 0.86
- **mlcore/ane_encoder (Apple Neural Engine Template)**（需要安装后验证）：Encoder-only transformer template verified to run 100% on Apple Neural Engine via MLComputePlan runtime inspection. Four constraints: FP16 precision, (B,C,1,S) 4D layout, ANE-allowlist operators, static shapes. Serves as recipe template for ESM2-150M on-device protein encoding. 证据：`mlcore/ane_encoder/README.md`, `mlcore/ane_encoder/README.md`, `mlcore/ane_encoder/README.md` Claim：`clm_0009` supported 0.86
- **Skill Authoring Framework**（可做安装前预览）：SKILL.md frontmatter + body format for domain knowledge skills consumed by the orbital router. Frontmatter includes name, description, keywords, and orb_class classification. 证据：`example-skills/hello-world/SKILL.md`, `skills/agent-loop/SKILL.md`, `skills/analytics/SKILL.md`, `skills/api/SKILL.md` Claim：`clm_0010` supported 0.86

## 怎么开始

- `npm install -g meridian-orbital` 证据：`README.md` Claim：`clm_0013` supported 0.86
- `claude mcp add meridian meridian-mcp` 证据：`README.md` Claim：`clm_0014` supported 0.86
- `npx -y meridian-orbital meridian-mcp-http` 证据：`README.md` Claim：`clm_0015` 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：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_0011` 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：`example-skills/hello-world/SKILL.md`, `example-skills/http-best-practices/SKILL.md`, `example-skills/markdown-formatting/SKILL.md`, `skills/agent-loop/SKILL.md` 等 Claim：`clm_0012` supported 0.86
- **能力存在：Orbital Task Routing (Meridian MCP)**（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`, `package.json` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86
- **能力存在：Dual-Transport MCP Server**（supported）：可以相信项目包含这类能力线索；是否适合你的具体任务仍要试用或安装后验证。 证据：`README.md`, `README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86
- **能力存在：Finance MCP (Binance/ Mercado Pago/ Coinbase)**（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：`finance-mcp/README.md`, `finance-mcp/README.md`, `finance-mcp/README.md`, `finance-mcp/package.json` Claim：`clm_0003` supported 0.86
- **能力存在：Pharmacy MCP (Farmacias del Pueblo VTEX)**（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：`pharmacy-mcp/README.md`, `pharmacy-mcp/README.md`, `pharmacy-mcp/package.json` Claim：`clm_0004` 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：`example-skills/hello-world/SKILL.md`, `example-skills/http-best-practices/SKILL.md`, `example-skills/markdown-formatting/SKILL.md`, `skills/agent-loop/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。

### 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：`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：`example-skills/hello-world/SKILL.md`, `example-skills/http-best-practices/SKILL.md`, `example-skills/markdown-formatting/SKILL.md`, `skills/agent-loop/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`, `binance-proxy/README.md`, `binance-proxy/package.json`, `finance-mcp/README.md` 等
- **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：`CHANGELOG.md`, `README.md`, `skills/docker-registry/SKILL.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**：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_0016` 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_0017` supported 0.86
- **风险**： 处理方式：
- **风险**： 处理方式：
- **风险**： 处理方式：
- **风险**： 处理方式：
- **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

- **Orbital Task Routing (Meridian MCP)**：先基于 role_skill_index / evidence_index 帮用户挑选可用角色、Skill 或工作流。 边界：可做安装前 Prompt 体验。 证据：`README.md`, `README.md`, `package.json` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86
- **Dual-Transport MCP Server**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`README.md`, `README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86
- **Finance MCP (Binance/ Mercado Pago/ Coinbase)**：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：`finance-mcp/README.md`, `finance-mcp/README.md`, `finance-mcp/README.md`, `finance-mcp/package.json` Claim：`clm_0003` supported 0.86
- **Pharmacy MCP (Farmacias del Pueblo VTEX)**：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：`pharmacy-mcp/README.md`, `pharmacy-mcp/README.md`, `pharmacy-mcp/package.json` Claim：`clm_0004` supported 0.86
- **Binance Proxy (Bright Data Tunnel)**：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：`binance-proxy/README.md`, `binance-proxy/README.md`, `binance-proxy/package.json` Claim：`clm_0005` supported 0.86
- **Helio Mirror (Solar Irradiance Triangulator)**：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：`helio-mirror/README.md`, `helio-mirror/README.md` Claim：`clm_0006` supported 0.86
- **Photon Route (Continuous-Variable Photonic Retrieval)**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`photon-route/README.md`, `photon-route/README.md` Claim：`clm_0007` supported 0.86
- **Design System / UI Kits**：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：`design-system/README.md`, `design-system/README.md`, `design-system/ui_kits/docs/README.md`, `design-system/ui_kits/miniapp/README.md` Claim：`clm_0008` supported 0.86
- **mlcore/ane_encoder (Apple Neural Engine Template)**：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：`mlcore/ane_encoder/README.md`, `mlcore/ane_encoder/README.md`, `mlcore/ane_encoder/README.md` Claim：`clm_0009` supported 0.86
- **Skill Authoring 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：`example-skills/hello-world/SKILL.md`, `skills/agent-loop/SKILL.md`, `skills/analytics/SKILL.md`, `skills/api/SKILL.md` Claim：`clm_0010` supported 0.86

### 上下文规模

- 文件总数：693
- 重要文件覆盖：40/693
- 证据索引条目：112
- 角色 / Skill 条目：90

### 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 meridian-orbital 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 meridian-orbital 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 meridian-orbital 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 90 个角色 / Skill / 项目文档条目。

- **hello-world**（skill）：A minimal example skill demonstrating the SKILL.md frontmatter + body format. Use as a template when authoring your own skills. 激活提示：当用户任务与“hello-world”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：`example-skills/hello-world/SKILL.md`
- **http-best-practices**（skill）：Patterns for HTTP clients + servers — idempotency, retries, timeouts, status code semantics, and request/response tracing. 激活提示：当用户任务与“http-best-practices”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：`example-skills/http-best-practices/SKILL.md`
- **markdown-formatting**（skill）：Clean, consistent markdown output for LLM-generated content — headings, lists, code fences, tables, and front matter. 激活提示：当用户任务与“markdown-formatting”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：`example-skills/markdown-formatting/SKILL.md`
- **agent-loop**（skill）：Production Claude agent loop — Session/Harness/Registry/Tool abstraction, DRY RUN safety guard, APScheduler integration, dead-letter error handling, tool registry, and observability hooks for autonomous agent systems 激活提示：当用户任务与“agent-loop”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/agent-loop/SKILL.md`
- **analytics**（skill）：Growth analytics authority — GA4 event tracking, conversion funnels, attribution models, cohort analysis, A/B testing, Supabase analytics queries, and revenue reporting patterns 激活提示：当用户任务与“analytics”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/analytics/SKILL.md`
- **api-reference**（skill）：OpenAPI/AsyncAPI spec authoring, Swagger UI hosting, endpoint doc patterns, request/response examples, error catalogs, and SDK reference generation for REST and WebSocket APIs 激活提示：当用户任务与“api-reference”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/api-reference/SKILL.md`
- **api**（skill）：FastAPI + async/sync HTTP client patterns, JWT auth, multi-provider routing, Pydantic request/response models, CORS, background tasks, and typed frontend API clients — synthesized from lead-gen-engine and seo-geo-aeo-engine 激活提示：当用户任务与“api”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/api/SKILL.md`
- **arxiv-taxonomy**（skill）：Complete arXiv category taxonomy, cross-listing rules, submission conventions, and query patterns. Use when fetching, filtering, or routing papers by discipline; when a user references "astro-ph.CO" or "hep-ph" and you need to know what that covers; when building recommender logic across adjacent categories. 激活提示：当用户任务与“arxiv-taxonomy”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/arxiv-taxonomy/SKILL.md`
- **astro-expert**（skill）：Astro framework expertise — islands architecture, SSR/hybrid modes, React integration, performance, and production-ready frontend scaffolding 激活提示：当用户任务与“astro-expert”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/astro-expert/SKILL.md`
- **auth**（skill）：JWT patterns, API key auth, Supabase Auth, OAuth2 flows, RBAC, FastAPI security dependencies 激活提示：当用户任务与“auth”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/auth/SKILL.md`
- **background-tasks**（skill）：Async task execution authority — Celery workers and beat scheduler, APScheduler in-process cron, bare Redis LPUSH/BRPOP queues, dead-letter queues, idempotency patterns, task monitoring, and Docker Compose worker definitions for Python pipelines 激活提示：当用户任务与“background-tasks”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/background-tasks/SKILL.md`
- **brightdata-collection**（skill）：Bright Data collection specialist — dynamic page retrieval, geo-targeted browsing, anti-bot resilient collection, SERP capture, and hard-target fallback for agent research pipelines 激活提示：当用户任务与“brightdata-collection”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/brightdata-collection/SKILL.md`
- **browser-stealth**（skill）：Browser automation stealth and residential proxy routing — patchright anti-detection, Playwright proxy at correct context level, sticky residential sessions, navigator.webdriver masking, datacenter vs residential IP fingerprinting, and page.evaluate fetch fallback 激活提示：当用户任务与“browser-stealth”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/browser-stealth/SKILL.md`
- **caching**（skill）：Redis patterns pub/sub, queues, TTL , in-process caching, cache-aside strategy, invalidation, Python and Node clients 激活提示：当用户任务与“caching”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/caching/SKILL.md`
- **causal-inference**（skill）：Causal inference for data scientists and analysts — DAGs and do-calculus, propensity score methods, difference-in-differences, instrumental variables, regression discontinuity, synthetic control, and variance reduction techniques CUPED , using CausalML, DoWhy, and rigorous A/B test analysis. 激活提示：当用户任务与“causal-inference”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/causal-inference/SKILL.md`
- **computer-vision**（skill）：Computer vision engineering covering object detection YOLO, RT-DETR , instance and semantic segmentation SAM, Mask R-CNN , image classification, OCR Tesseract, PaddleOCR , video understanding, OpenCV pipelines, PyTorch/torchvision, ONNX export, and edge deployment via TensorRT and CoreML. 激活提示：当用户任务与“computer-vision”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/computer-vision/SKILL.md`
- **container-security**（skill）：Container runtime security — non-root users, read-only filesystems, capability dropping, secrets hygiene, image scanning, supply-chain verification, and runtime policy enforcement 激活提示：当用户任务与“container-security”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/container-security/SKILL.md`
- **content**（skill）：Content strategy authority — editorial calendar, content briefs, topic clustering, AI-assisted writing workflows, UGC patterns, publishing pipelines, and brand voice guidelines across blog, social, and email formats 激活提示：当用户任务与“content”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/content/SKILL.md`
- **context-compaction**（skill）：Reduce context entropy and token burn by compacting stale state, preserving only reusable signal, and reshaping long tasks into short high-coherence context windows 激活提示：当用户任务与“context-compaction”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/context-compaction/SKILL.md`
- **css-spacing-layout**（skill）：CSS Spacing & Container Layout Architect — expert in padding, margin, borders, container hierarchy, spacing scales, and layout consistency across breakpoints 激活提示：当用户任务与“css-spacing-layout”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/css-spacing-layout/SKILL.md`
- **curl-recipes**（skill）：curl one-liners for HTTP debugging, API testing, auth header patterns, file upload, response inspection, timing breakdown, and connection tracing 激活提示：当用户任务与“curl-recipes”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/curl-recipes/SKILL.md`
- **database**（skill）：Relational database authority — PostgreSQL, async ORM SQLAlchemy 2 / asyncpg , Alembic schema migrations, Supabase Python and JS clients, query optimisation, index strategy, connection pooling, transaction patterns, and bulk operations across Python and Node stacks 激活提示：当用户任务与“database”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/database/SKILL.md`
- **defi-protocols**（skill）：DeFi protocol authority — flashloans, AMM design, lending markets, concentrated liquidity, MEV, triangular and cross-DEX arbitrage, stablecoin depeg dynamics, and on-chain capital routing across Uniswap V3, Aerodrome, Balancer, Curve, and Aave V3 on Ethereum, Base, Arbitrum, Optimism, and Polygon 激活提示：当用户任务与“defi-protocols”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/defi-protocols/SKILL.md`
- **devops**（skill）：General-purpose DevOps for VPS-based deployments — Nginx reverse proxy, SSL with Let's Encrypt, PM2 process management, systemd services, firewall rules, log rotation, monitoring, and backup strategies 激活提示：当用户任务与“devops”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/devops/SKILL.md`
- **dns**（skill）：DNS record management, Cloudflare API automation, TTL strategy, Let's Encrypt DNS challenges, split-horizon DNS, and resolver debugging for production domains 激活提示：当用户任务与“dns”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/dns/SKILL.md`
- **doc-templates**（skill）：Reusable documentation fragments and micro-patterns — ADR templates, runbook blocks, incident post-mortem skeletons, PR description stubs, and changelog entry formats 激活提示：当用户任务与“doc-templates”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/doc-templates/SKILL.md`
- **docker-compose**（skill）：Multi-service docker-compose orchestration — service dependency ordering, profiles, override files, environment-specific configs, rolling restarts, and compose-based dev/prod parity patterns 激活提示：当用户任务与“docker-compose”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/docker-compose/SKILL.md`
- **docker-registry**（skill）：Container image registry workflows — GHCR, Docker Hub, and private registry auth, tagging strategies, CI push pipelines, image pruning, and multi-platform manifest publishing 激活提示：当用户任务与“docker-registry”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/docker-registry/SKILL.md`
- **docker**（skill）：Docker platform authority — image architecture, multi-stage builds, runtime security, networking models, volume strategies, container orchestration patterns, registry workflows, and production-grade containerization across Python, Node, and multi-service systems 激活提示：当用户任务与“docker”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/docker/SKILL.md`
- **docs-strategy**（skill）：Documentation architecture, information hierarchy, style guides, versioning strategy, and cross-team documentation governance for production software projects 激活提示：当用户任务与“docs-strategy”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/docs-strategy/SKILL.md`
- **docs-update**（skill）：Rapidly catch up on any codebase by reading only .md files first — orientation protocol, what each doc type reveals, how to ask targeted follow-up questions, and how to keep docs in sync with code changes 激活提示：当用户任务与“docs-update”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/docs-update/SKILL.md`
- **edward-sturm**（skill）：Partner skill simulating Edward Sturm's SEO, GEO, and AEO methodology — evidence-backed reasoning patterns drawn exclusively from verified source corpus 激活提示：当用户任务与“edward-sturm”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/edward-sturm/SKILL.md`
- **env-management**（skill）：Environment variable patterns for Python python-dotenv, Pydantic BaseSettings and TypeScript VITE , NEXT PUBLIC projects — .env.example authorship, schema validation, mock/feature flags, and safe defaults — synthesized from lead-gen-engine and seo-geo-aeo-engine 激活提示：当用户任务与“env-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：`skills/env-management/SKILL.md`
- **event-driven**（skill）：Event-driven architecture authority — Redis pub/sub, event bus patterns, async event pipelines, channel management, startup recovery, dead-letter handling, and reactive system design 激活提示：当用户任务与“event-driven”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/event-driven/SKILL.md`
- **exa-search**（skill）：Exa search specialist — semantic web discovery, source ranking, entity finding, citation set building, and lightweight content retrieval for evidence-driven agent workflows 激活提示：当用户任务与“exa-search”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/exa-search/SKILL.md`
- **federated-learning**（skill）：Federated learning engineering covering FedAvg and FedProx aggregation, differential privacy DP-SGD, Opacus , secure aggregation protocols, PySyft, the Flower framework, split learning, on-device training, communication compression, non-IID data heterogeneity, and model poisoning defenses for privacy-preserving distributed ML. 激活提示：当用户任务与“federated-learning”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/federated-learning/SKILL.md`
- **fine-tuning**（skill）：LLM fine-tuning authority — LoRA, QLoRA, and full fine-tuning workflows with PEFT, Axolotl, and Unsloth; supervised fine-tuning SFT , DPO, and RLHF alignment; dataset curation and formatting; GPTQ/AWQ quantization; vLLM serving; and evaluation with lm-evaluation-harness 激活提示：当用户任务与“fine-tuning”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/fine-tuning/SKILL.md`
- **firecrawl-extract**（skill）：Firecrawl extraction specialist — scrape known URLs into clean markdown, map sites, crawl documentation and blog surfaces, and normalize content capture for downstream reasoning or indexing 激活提示：当用户任务与“firecrawl-extract”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/firecrawl-extract/SKILL.md`
- **firewall**（skill）：Production firewall management — UFW policy design, iptables/nftables rule authoring, stateful connection tracking, ingress/egress filtering, fail2ban integration, and firewall audit workflows for Linux servers 激活提示：当用户任务与“firewall”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/firewall/SKILL.md`
- **Framer Motion**（skill）：Framer Motion animation library — setup, motion principles, staggered lists, exit transitions, shared layout, gesture feedback, accessibility 激活提示：当用户任务与“Framer Motion”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/framer-motion/SKILL.md`
- **github**（skill）：GitHub workflow — creating and managing repos, always working on feature/development branches never committing core changes directly to main , PR conventions, branch protection, and gh CLI patterns 激活提示：当用户任务与“github”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/github/SKILL.md`
- **internal-docs-operator**（skill）：Use when working against the internal documentation surfaces on :4401 and :4400, especially for target-specific doc updates, source-of-truth checks, verification of live state, and preventing false completion claims 激活提示：当用户任务与“internal-docs-operator”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/internal-docs-operator/SKILL.md`
- **ip-blocklist**（skill）：IP allowlist and blocklist fragments — iptables/nftables set-based blocking, nginx geo blocks, Cloudflare IP rules, and dynamic blocklist management for network access control 激活提示：当用户任务与“ip-blocklist”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/ip-blocklist/SKILL.md`
- **knowledge**（skill）：Knowledge systems authority — RAG pipelines, vector search with Qdrant, embedding generation, semantic chunking, knowledge graph construction, retrieval evaluation, and citation-aware QA patterns 激活提示：当用户任务与“knowledge”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/knowledge/SKILL.md`
- **kubernetes**（skill）：Kubernetes orchestration authority — pods, deployments, services, ingress, helm charts, kustomize overlays, RBAC, network policies, operators, StatefulSets for databases, production-grade cluster hardening, and kubectl workflows 激活提示：当用户任务与“kubernetes”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/kubernetes/SKILL.md`
- **latex**（skill）：LaTeX and mathematical notation authority — typesetting equations, document structure, AMS math environments, BibTeX citations, arXiv submission formatting, KaTeX/MathJax web rendering, and converting mathematical prose to precise symbolic notation 激活提示：当用户任务与“latex”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/latex/SKILL.md`
- **llm-integration**（skill）：Claude and OpenAI API patterns, prompt design, context management, streaming, tool use, cost control, multi-model routing 激活提示：当用户任务与“llm-integration”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/llm-integration/SKILL.md`
- **ml-training**（skill）：Machine learning model training authority — classifier and regressor training with scikit-learn and PyTorch, feature engineering, cross-validation, hyperparameter tuning, fine-tuning transformer models with HuggingFace, dataset splits, loss functions, learning rate schedules, and reproducible training runs 激活提示：当用户任务与“ml-training”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/ml-training/SKILL.md`
- **mlops**（skill）：MLOps infrastructure engineering covering model serving vLLM, TorchServe, Triton , feature stores, model registries MLflow, W&B , A/B and shadow deployments, drift detection, Prometheus/Grafana ML metrics, Kubeflow Pipelines, and Airflow DAGs for end-to-end training and deployment pipelines. 激活提示：当用户任务与“mlops”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/mlops/SKILL.md`
- **multi-agent-collab**（skill）：Running multiple AI coding agents Claude Code, Codex, OpenClaw in isolated containers that share a single repo clone via /opt bind mounts, communicate over SSH, and avoid redundant git clones — container topology, volume strategy, inter-agent SSH, and coordination patterns 激活提示：当用户任务与“multi-agent-collab”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/multi-agent-collab/SKILL.md`
- **multi-llm-routing**（skill）：Multi-LLM routing authority — provider selection, cost-aware tiering, fallback chains, latency budgets, health checks, A/B model testing, and unified client abstraction across Anthropic/OpenAI/Mistral 激活提示：当用户任务与“multi-llm-routing”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/multi-llm-routing/SKILL.md`
- **network**（skill）：Framework-agnostic production networking authority — TCP/IP stack, interface and routing management, firewall policy, traffic inspection, VPN, bandwidth diagnostics, and cross-layer reachability debugging for bare-metal and cloud VPS environments 激活提示：当用户任务与“network”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/network/SKILL.md`
- **nextjs-dashboard**（skill）：Next.js 15+ App Router dashboard skill — server/client components, mobile nav, polling, data tables, skeleton states, and production-grade dashboard UX patterns 激活提示：当用户任务与“nextjs-dashboard”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/nextjs-dashboard/SKILL.md`
- **observability**（skill）：Structured logging structlog/Python, pino/Node , health endpoints, PM2 and Docker metrics, alerting patterns 激活提示：当用户任务与“observability”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/observability/SKILL.md`
- **orchestration**（skill）：Multi-agent orchestration authority — agent memory protocols, task delegation, SOUL/AGENTS patterns, session startup, inter-agent communication, OpenClaw workspace coordination, and agent lifecycle management 激活提示：当用户任务与“orchestration”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/orchestration/SKILL.md`
- **outbound**（skill）：Outbound growth authority — cold email sequences, lead scoring, ICP definition, CRM integration, Apollo and Instantly patterns, deliverability, reply handling, and pipeline-to-revenue tracking 激活提示：当用户任务与“outbound”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/outbound/SKILL.md`
- **partner-skill-compiler**（skill）：Partner skill compiler authority — synthesize grounded public evidence into a reusable partner prompt, persona profile, and SKILL.md with worldview, operating style, audience, offers, voice anchors, and strict anti-hallucination constraints 激活提示：当用户任务与“partner-skill-compiler”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/partner-skill-compiler/SKILL.md`
- **persona-research**（skill）：Public persona research authority — discover, rank, and curate public sources for a person across websites, YouTube, podcasts, newsletters, LinkedIn, and social profiles while avoiding name collisions, wrong-identity contamination, and low-signal pages 激活提示：当用户任务与“persona-research”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/persona-research/SKILL.md`
- **physics-units-si**（skill）：SI units, CODATA constants, dimensional analysis, and rigorous unit conversion. Use when extracting, normalizing, or computing physical quantities energy, frequency, length, mass, time, temperature . Handles eV/keV/MeV/GeV/TeV, Hz/kHz/MHz/GHz/THz, meter/cm/Å/fm, kg/g/u, second/ms/μs/ns/fs, Kelvin/eV-temperature equivalents. 激活提示：当用户任务与“physics-units-si”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/physics-units-si/SKILL.md`
- **port-scanner**（skill）：nmap and masscan recipes for host discovery, open port enumeration, service fingerprinting, and firewall validation on VPS deployments 激活提示：当用户任务与“port-scanner”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/port-scanner/SKILL.md`
- **prompt-engineering**（skill）：Prompt engineering planet — system prompt architecture, few-shot and chain-of-thought design, ReAct and tool-use patterns, context window management, structured output, temperature and sampling tuning, jailbreak resistance, and model-specific prompt differences across Claude, GPT-4, Gemini, and open-weight models 激活提示：当用户任务与“prompt-engineering”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/prompt-engineering/SKILL.md`
- **rate-limiting**（skill）：Rate limiting fragments — nginx request throttling, iptables connection rate rules, fail2ban threshold tuning, and per-IP burst control for API and web services 激活提示：当用户任务与“rate-limiting”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/rate-limiting/SKILL.md`
- **react dashboard cards**（skill）：Production-grade React dashboard card patterns — StatCard, KPI strip, domain progress cards, metric bars, badge system, skeleton states, glass-card CSS variables. Extracted from lead-gen and seo-geo-aeo dashboards. 激活提示：当用户任务与“react dashboard cards”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/react-dashboard-cards/SKILL.md`
- **react-native**（skill）：React Native mobile app authority — Expo SDK, native modules, iOS and Android builds, push notifications, deep linking, App Store and Play Store submission, over-the-air updates, native gesture handlers, and cross-platform mobile app architecture 激活提示：当用户任务与“react-native”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/react-native/SKILL.md`
- **reasoning**（skill）：AI reasoning authority — prompt engineering, chain-of-thought, self-consistency, evaluation frameworks, LLM judging, structured output, and prompt versioning patterns for production AI systems 激活提示：当用户任务与“reasoning”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/reasoning/SKILL.md`
- **schema-authority**（skill）：Single source of truth authority — define once in Pydantic / SQLModel / OpenAPI / Protobuf, generate TypeScript types, validators, docs, mocks, and contract tests from that canonical definition; schema registry patterns, spec-first API design, consumer-driven contract testing, and database-model alignment 激活提示：当用户任务与“schema-authority”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/schema-authority/SKILL.md`
- **scientific-writing-voice**（skill）：Voice and structure rules for science journalism prose — the register used by Quanta Magazine, Nature News, New Scientist research features, and high-caliber popular-science briefs. Use when summarizing a technical paper for a physics-literate but time-constrained reader, drafting research digests, or writing a "brief" that preserves quantitative precision without jargon walls. 激活提示：当用户任务与“scientific-writing-voice”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/scientific-writing-voice/SKILL.md`
- **security-audit**（skill）：Smart contract security auditing authority — reentrancy, flash loan attacks, oracle manipulation, access control vulnerabilities, integer overflow/underflow, front-running, signature replay, proxy storage collisions, static analysis with Slither and Mythril, fuzz testing with Echidna, and professional audit report writing 激活提示：当用户任务与“security-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：`skills/security-audit/SKILL.md`
- **semantic-scholar-api**（skill）：Semantic Scholar + OpenAlex citation graph APIs for walking paper-to-paper relationships, finding citers/references, author disambiguation, and discovering related work. Use when arxiv or keyword search is insufficient — e.g. "what papers cite this one", "who else works on X", "what's the precursor work behind this abstract". Both APIs are free; no key required for reasonable usage. 激活提示：当用户任务与“semantic-scholar-api”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/semantic-scholar-api/SKILL.md`
- **seo**（skill）：SEO authority — SERP analysis, GEO/AEO citation patterns, E-E-A-T signals, keyword research, technical SEO audits, structured data, and content-to-ranking pipeline patterns for AI-assisted publishing workflows 激活提示：当用户任务与“seo”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/seo/SKILL.md`
- **shadcn/ui expert**（skill）：shadcn/ui composable component system — setup, token strategy, card patterns, form composition, real project examples from lead-gen dashboard 激活提示：当用户任务与“shadcn/ui expert”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/shadcn-ui/SKILL.md`
- **solidity**（skill）：Solidity smart-contract authoring authority — ERC20/ERC721/ERC1155 token standards, reentrancy guards, access control, gas optimisation, upgradeable proxies, Foundry testing, fuzzing, invariant checks, and production contract security for EVM chains 激活提示：当用户任务与“solidity”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/solidity/SKILL.md`
- **ssh-hardening**（skill）：SSH daemon hardening fragments — sshd config directives, key-only auth, port and user restrictions, agent forwarding policy, and authorized keys access control patterns 激活提示：当用户任务与“ssh-hardening”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/ssh-hardening/SKILL.md`
- **ssl-tls**（skill）：TLS certificate lifecycle — Let's Encrypt HTTP-01 and DNS-01 issuance, renewal automation, mTLS, HSTS, OCSP stapling, certificate inspection, and Nginx TLS hardening 激活提示：当用户任务与“ssl-tls”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/ssl-tls/SKILL.md`
- **state-management**（skill）：Frontend single source of truth — server state with TanStack Query caching, invalidation, optimistic updates , client state with Zustand slices, middleware, persistence , URL as state, Redux Toolkit for coordinated entity state, and the discipline of never duplicating remote data in local stores 激活提示：当用户任务与“state-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：`skills/state-management/SKILL.md`
- **tailwind-css expert**（skill）：Tailwind CSS — configuration, token strategy, CSS variable theming, color-mix patterns, button variants, card utilities, dark mode, production quality checklist 激活提示：当用户任务与“tailwind-css expert”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/tailwind-css/SKILL.md`
- **token-awareness**（skill）：Token cost–benefit and energy analysis before implementing — estimate read/write overhead, compare full implementation vs targeted workaround, model signal-per-token efficiency, and apply the minimum-effective-change principle 激活提示：当用户任务与“token-awareness”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/token-awareness/SKILL.md`
- **token-economics**（skill）：Model token usage as an energy budget — estimate workload cost, compare throughput per watt / per dollar, and choose routes, models, and architectures with the best signal-per-token ratio 激活提示：当用户任务与“token-economics”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/token-economics/SKILL.md`
- **tokenomics**（skill）：Token economic design authority — supply schedules, emission curves, vesting and cliff structures, bonding curves, liquidity bootstrapping pools, governance token design, veToken and vote-escrowed models, staking reward mechanics, token sink design, incentive alignment analysis, and Gini coefficient distribution analysis for crypto protocols 激活提示：当用户任务与“tokenomics”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/tokenomics/SKILL.md`
- **transcript-ingestion**（skill）：Transcript ingestion authority — capture, normalize, and prioritize transcripts from YouTube videos, podcast pages, captions, show notes, and transcript-like page content for persona systems, RAG pipelines, and grounded synthesis 激活提示：当用户任务与“transcript-ingestion”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/transcript-ingestion/SKILL.md`
- 其余 10 个条目见 `AI_CONTEXT_PACK.json`。

## 证据索引

- 共索引 112 条证据。

- **Docs UI kit — ask-meridian.uk/docs**（documentation）：Documentation surface for Meridian. The third top-level kit alongside landing/ and miniapp/ . 证据：`design-system/ui_kits/docs/README.md`
- **Meridian MCP**（documentation）：Dynamic task routing via orbital mechanics. Domain-agnostic — candidates can be tools, prompts, documents, products, or any routable entity. 证据：`README.md`
- **binance-proxy**（documentation）：Tiny HTTP proxy that bridges a CF Worker finance-mcp to Binance via Bright Data. Cloudflare Workers' egress IPs are geo-blocked by Binance, and CF Workers' cloudflare:sockets.startTls can't override SNI after a CONNECT, so this thin Fly machine handles the BD CONNECT-tunnel + TLS upgrade on the Worker's behalf. Bright Data hands us a single static residential-grade exit IP that we whitelist on the Binance API key. 证据：`binance-proxy/README.md`
- **Meridian Design System**（documentation）：Canonical reference for Meridian's visual language: tokens, type scale, components, motion, and brand assets. The tokens here mirror the values already in landing/style.css and miniapp/miniapp.css — this directory pulls them into one importable surface and adds full reference kits for each app. 证据：`design-system/README.md`
- **finance-mcp**（documentation）：Personal finance MCP that runs on Cloudflare Workers. Single-tenant. Passkey-bootstrapped: you register a passkey once via a one-time admin link, and from then on adding the connector to Grok / Claude.ai / ChatGPT is just OAuth + a passkey tap. 证据：`finance-mcp/README.md`
- **helio-mirror**（documentation）：Multi-spacecraft solar irradiance triangulator. 证据：`helio-mirror/README.md`
- **pharmacy-mcp**（documentation）：MCP for Farmacias del Pueblo Argentine VTEX storefront . Lets your AI host search the catalog, build a cart over time, and hand the cart off to your browser for the final login + payment. 证据：`pharmacy-mcp/README.md`
- **photon-route**（documentation）：A continuous-variable CV photonic retrieval research artifact. 证据：`photon-route/README.md`
- **Sim reports**（documentation）：Automated artifacts from the recurring simulation workflows: 证据：`data/sim-reports/README.md`
- **Sign in with Apple — official brand assets**（documentation）：Sign in with Apple — official brand assets 证据：`design-system/auth/sign-in-with-apple/README.md`
- **Helio UI kit — solar irradiance dashboard**（documentation）：Helio UI kit — solar irradiance dashboard 证据：`design-system/ui_kits/helio/README.md`
- **Helix UI kit — proteins-as-star-systems**（documentation）：Helix UI kit — proteins-as-star-systems 证据：`design-system/ui_kits/helix/README.md`
- **Landing UI kit — ask-meridian.uk**（documentation）：Marketing surface for Meridian. Recreated from landing/style.css + landing/index.html in LuuOW/meridian-mcp . Includes the full hero with orbital SVG diagram, fleet grid, stats strip, features, how-it-works, pricing, and the final CTA. 证据：`design-system/ui_kits/landing/README.md`
- **Lens UI kit — WebXR Vision Lab**（documentation）：The headset-side surface. Vision-language model in browser, candidates orbiting in real space. Point a controller at an object, pull the trigger, the model describes what it sees. 证据：`design-system/ui_kits/lens/README.md`
- **Miniapp UI kit — Task Orbit**（documentation）：The flagship Meridian demo: type a task, get five ranked candidates with celestial classes, click one to see why. Hosted at ask-meridian.uk/miniapp . 证据：`design-system/ui_kits/miniapp/README.md`
- **Photon Router UI kit — Strawberry Fields retrieval**（documentation）：Photon Router UI kit — Strawberry Fields retrieval 证据：`design-system/ui_kits/photon-router/README.md`
- **Third-party brand assets**（documentation）：This directory holds trademarked brand marks owned by their respective companies , included to identify compatibility e.g. "works with Grok" on the Meridian site. 证据：`landing/img/brand/README.md`
- **mlcore/ane encoder**（documentation）：A small encoder-only transformer, written in the Apple Neural Engine idiom from day one , converted to a CoreML MLPackage, and verified to land 100% on the Apple Neural Engine at runtime — confirmed programmatically via MLComputePlan , not just Xcode GUI inspection. 证据：`mlcore/ane_encoder/README.md`
- **Package**（package_manifest）：{ "name": "binance-proxy", "version": "1.1.0", "type": "module", "private": true, "engines": { "node": " =20" }, "scripts": { "start": "node index.mjs" }, "dependencies": { "https-proxy-agent": "^7.0.6" } } 证据：`binance-proxy/package.json`
- **Package**（package_manifest）：{ "name": "finance-mcp", "version": "0.2.0", "description": "Personal finance MCP — passkey-bootstrapped OAuth, passwordless after setup.", "type": "module", "private": true, "scripts": { "dev": "wrangler dev", "deploy": "wrangler deploy", "typecheck": "tsc --noEmit", "create-link": "node scripts/create-link.mjs" }, "dependencies": { "@simplewebauthn/server": "^11.0.0" }, "devDependencies": { "@cloudflare/workers-types": "^4.20240909.0", "typescript": "^5.5.0", "wrangler": "^3.78.0" } } 证据：`finance-mcp/package.json`
- **Package**（package_manifest）：{ "name": "meridian-orbital", "version": "3.2.0", "description": "Self-contained MCP — orbital task router with online-learning classifier stdio + HTTP/Streamable transports — Grok / ChatGPT / Claude compatible . Generates candidates with Llama-3.3-70B via GitHub Models and ranks them via a local orbital classifier planet/moon/trojan/asteroid/comet/irregular . Domain-agnostic — candidates can be tools, prompts, documents, products. No backend required.", "type": "module", "bin": { "meridian-mcp": "./mcp/index.mjs", "meridian-mcp-http": "./mcp/http.mjs" }, "main": "./mcp/index.mjs", "files": "mcp", "README.md", "LICENSE" , "scripts": { "stdio": "node mcp/index.mjs", "http": "node mcp/http.mj… 证据：`package.json`
- **Package**（package_manifest）：{ "name": "pharmacy-mcp", "version": "0.1.0", "private": true, "type": "module", "description": "MCP for Farmacias del Pueblo VTEX . Browser-mediated reCAPTCHA bootstrap; headless cart automation thereafter.", "scripts": { "dev": "wrangler dev", "deploy": "wrangler deploy", "tail": "wrangler tail" }, "dependencies": { "@modelcontextprotocol/sdk": "^1.0.4" }, "devDependencies": { "wrangler": "^3.99.0" } } 证据：`pharmacy-mcp/package.json`
- **hello-world**（skill_instruction）：This is the body of the skill. The orbital router reads the frontmatter description to compute relevance to a task. 证据：`example-skills/hello-world/SKILL.md`
- **http-best-practices**（skill_instruction）：Timeouts - Always set connect timeout ≤5s and read timeout ≤30s - Never trust defaults; they vary by library 证据：`example-skills/http-best-practices/SKILL.md`
- **markdown-formatting**（skill_instruction）：Heading hierarchy - Use only one H1 per document - H2 for major sections, H3 for subsections; never skip levels 证据：`example-skills/markdown-formatting/SKILL.md`
- **agent-loop**（skill_instruction）：Production patterns for building autonomous agent loops with the Anthropic API. Covers the Session/Harness/Registry/Tool abstraction, safe development mode via DRY RUN, APScheduler cron integration, and observability hooks. Designed for systems that run unsupervised on a schedule and must never silently bill or silently fail. 证据：`skills/agent-loop/SKILL.md`
- **analytics**（skill_instruction）：Covers measurement, attribution, and reporting for growth systems: web analytics, conversion funnels, cohort analysis, and revenue tracking. 证据：`skills/analytics/SKILL.md`
- **api-reference**（skill_instruction）：Production patterns for writing, validating, and publishing machine-readable API reference documentation. Covers OpenAPI 3.1, AsyncAPI 2.x, doc generation from code annotations, and consumer-facing reference portals. 证据：`skills/api-reference/SKILL.md`
- **api**（skill_instruction）：Production patterns for building and consuming APIs in the Python/FastAPI + TypeScript/React stack used across lead-gen-engine and seo-geo-aeo-engine. 证据：`skills/api/SKILL.md`
- **arxiv-taxonomy**（skill_instruction）：When to invoke - Building an arXiv query wildcards, cross-listings, OR combinations - Interpreting a paper's primary category vs secondary - Recommending adjacent categories to a user interested in X - Explaining to a user why a paper on gravitational waves shows up in both astro-ph.HE and gr-qc 证据：`skills/arxiv-taxonomy/SKILL.md`
- **astro-expert**（skill_instruction）：Expert knowledge for building production-ready frontends with Astro — combining static rendering, selective hydration, and React islands for interactive components. 证据：`skills/astro-expert/SKILL.md`
- **auth**（skill_instruction）：Authentication and authorisation patterns across the stack: JWTs, API keys, Supabase Auth, and role-based access control. 证据：`skills/auth/SKILL.md`
- **background-tasks**（skill_instruction）：Covers async task execution: Celery + beat for scheduled/queued work, APScheduler for in-process cron, and bare Redis queues for lightweight pipelines. 证据：`skills/background-tasks/SKILL.md`
- **brightdata-collection**（skill_instruction）：Use this only when simpler search or extraction paths are insufficient. 证据：`skills/brightdata-collection/SKILL.md`
- **browser-stealth**（skill_instruction）：Patterns for running Playwright via patchright through a residential proxy without triggering bot detection. Covers the critical proxy configuration level bug, sticky session IDs, webdriver masking, and the page.evaluate fetch fallback when context.request can't authenticate through the proxy. 证据：`skills/browser-stealth/SKILL.md`
- **caching**（skill_instruction）：Covers Redis as cache, pub/sub bus, and job queue — plus lightweight in-process caching for Node/Python services. 证据：`skills/caching/SKILL.md`
- **Causal Inference**（skill_instruction）：Causal inference is the discipline of estimating cause-and-effect relationships from data — answering "what would have happened?" rather than "what is correlated with what?" This skill covers the identification strategies that make causal claims defensible DAGs, assumptions, estimation methods and the computational tools CausalML, DoWhy that implement them. It appears alongside analytics and ML engineering whenever the goal is decision-making, not prediction. 证据：`skills/causal-inference/SKILL.md`
- **Computer Vision**（skill_instruction）：Computer vision is the engineering discipline of making machines interpret and act on visual data — images, video, point clouds, and multi-modal streams. This skill spans the full pipeline from raw pixel ingestion through model training, evaluation, export, and production deployment, with emphasis on the practical trade-offs between accuracy, latency, and hardware constraints. Canonical frameworks are PyTorch/torchvision, Ultralytics, OpenCV, and the ONNX ecosystem. 证据：`skills/computer-vision/SKILL.md`
- **container-security**（skill_instruction）：Security hardening for containerized workloads at every layer: image build time, runtime configuration, secrets handling, and supply-chain verification. Applied in docker-compose, Kubernetes, or bare Docker. 证据：`skills/container-security/SKILL.md`
- **content**（skill_instruction）：Covers the full content lifecycle: strategy, briefing, production AI-assisted and human , and distribution. Applies across blog, email, UGC, and social formats. 证据：`skills/content/SKILL.md`
- **context-compaction**（skill_instruction）：Compaction is not summarization for its own sake. It is energy recovery. 证据：`skills/context-compaction/SKILL.md`
- **CSS Spacing & Container Layout Architect**（skill_instruction）：CSS Spacing & Container Layout Architect 证据：`skills/css-spacing-layout/SKILL.md`
- **curl-recipes**（skill_instruction）：Targeted curl patterns for debugging HTTP services, testing API endpoints, inspecting TLS, and measuring request timing. Micro-skill — composable with network, ssl-tls, api, and webhook debugging workflows. 证据：`skills/curl-recipes/SKILL.md`
- **database**（skill_instruction）：Authoritative reference for relational database work: connection pooling, async queries, migrations, and Supabase-specific patterns. 证据：`skills/database/SKILL.md`
- **defi-protocols**（skill_instruction）：Production DeFi routing across lending, AMMs, and arbitrage. Covers how capital moves on-chain: flashloan mechanics, AMM math, liquidity fragmentation, and the physics of price discovery across DEXes and chains. 证据：`skills/defi-protocols/SKILL.md`
- **devops**（skill_instruction）：Practical DevOps for self-hosted VPS deployments running Python/FastAPI + Node/Astro apps. Assumes Ubuntu 22.04/24.04, Docker optional. 证据：`skills/devops/SKILL.md`
- **dns**（skill_instruction）：DNS patterns for production domains managed through Cloudflare or direct registrar control. Covers record types, automation via API, debugging resolution chains, and DNS-01 challenges for wildcard TLS certificates. 证据：`skills/dns/SKILL.md`
- **doc-templates**（skill_instruction）：Copy-paste fragments for recurring documentation structures. Each template is self-contained and opinionated. Pick the block you need; don't combine templates for the same page. 证据：`skills/doc-templates/SKILL.md`
- **docker-compose**（skill_instruction）：Production-grade docker-compose.yml authoring for multi-service applications. Covers dependency graphs, conditional service activation, config layering, and zero-downtime update strategies. 证据：`skills/docker-compose/SKILL.md`
- **docker-registry**（skill_instruction）：Production patterns for pushing, pulling, tagging, and managing container images in registries. Covers GitHub Container Registry GHCR , Docker Hub, and self-hosted registries. 证据：`skills/docker-registry/SKILL.md`
- **docker**（skill_instruction）：Production containerization platform for Python/FastAPI, Node/Astro, and multi-service architectures. Covers the full lifecycle: image design, build pipelines, runtime configuration, security hardening, networking, storage, and operational observability. Independent of any specific cloud provider or orchestrator. 证据：`skills/docker/SKILL.md`
- **docs-strategy**（skill_instruction）：Authoritative patterns for building and maintaining documentation systems — from single-product READMEs to multi-tenant doc sites, living style guides, and versioned reference portals. 证据：`skills/docs-strategy/SKILL.md`
- **docs-update**（skill_instruction）：How to get productive in an unfamiliar or returning codebase as fast as possible by reading structured documentation before touching any code. Also covers keeping docs current so the next session costs fewer tokens. 证据：`skills/docs-update/SKILL.md`
- **Edward Sturm Partner Skill**（skill_instruction）：Skill type : Evidence-backed partner simulation Deployed by : partner-skill-compiler Target pipeline : seo-geo-aeo-engine Anti-hallucination boundary : This skill simulates reasoning patterns, methodology, and voice drawn exclusively from the verified source corpus below. It does not invent facts, statistics, client names, or claims not present in the corpus. When a question falls outside corpus coverage, the skill flags the gap explicitly rather than extrapolating. 证据：`skills/edward-sturm/SKILL.md`
- **env-management**（skill_instruction）：Patterns for managing environment variables across Python/FastAPI backends and TypeScript/Next.js or Vite frontends. Synthesized from lead-gen-engine python-dotenv + manual validation and seo-geo-aeo-engine Pydantic BaseSettings + LRU cache . 证据：`skills/env-management/SKILL.md`
- **event-driven**（skill_instruction）：Covers how to design, implement, and operate event-driven systems: pub/sub messaging, event buses, async pipelines, and reactive coordination between services. 证据：`skills/event-driven/SKILL.md`
- **exa-search**（skill_instruction）：Use this when the task begins with discovery rather than known URLs. 证据：`skills/exa-search/SKILL.md`
- **Federated Learning**（skill_instruction）：Federated learning FL trains machine learning models across decentralized data sources without centralizing raw data — devices or silos compute local updates, and only model parameters gradients or weights are aggregated by a coordinator. This skill covers the algorithmic foundations aggregation strategies, privacy guarantees, communication efficiency , the production frameworks Flower, PySyft , and the threat models poisoning, inference attacks that determine whether a FL deployment is actually privacy-preserving or merely privacy-theater. 证据：`skills/federated-learning/SKILL.md`
- **Fine-Tuning**（skill_instruction）：Production authority on adapting large language models: parameter-efficient fine-tuning with LoRA/QLoRA, full supervised fine-tuning, preference alignment via DPO and RLHF, quantization for deployment, and systematic evaluation. Use this skill when training or adapting any LLM beyond prompting, including domain adaptation, instruction following, and RLHF pipelines. 证据：`skills/fine-tuning/SKILL.md`
- **firecrawl-extract**（skill_instruction）：Use this when URLs are known and the agent needs clean content, crawl expansion, or sitemap-style discovery. 证据：`skills/firecrawl-extract/SKILL.md`
- 其余 52 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：`design-system/ui_kits/docs/README.md`, `README.md`, `binance-proxy/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：`design-system/ui_kits/docs/README.md`, `README.md`, `binance-proxy/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, package.json
- **轨道分类器详解**：importance `high`
  - source_paths: mcp/_lib/orbital.mjs, mcp/_lib/core.mjs, mcp/_lib/tokenize.mjs
- **系统架构**：importance `high`
  - source_paths: mcp/index.mjs, mcp/http.mjs, cf-worker/worker.mjs, api-worker/worker.mjs
- **数据流与反馈机制**：importance `high`
  - source_paths: cf-worker/online_learning.mjs, scripts/calibrate-classifier.mjs, scripts/calibration-baseline.json
- **部署指南**：importance `high`
  - source_paths: Dockerfile, binance-proxy/Dockerfile, cf-worker/wrangler.toml, api-worker/wrangler.toml
- **运维手册**：importance `medium`
  - source_paths: OPERATIONS.md, MONOREPO.md, scripts/audit.mjs, scripts/simulate-ci-methods.mjs
- **扩展与定制开发**：importance `medium`
  - source_paths: skills/skill_orbit.py, skills/agent-loop/SKILL.md, example-skills/hello-world/SKILL.md, scripts/new-skill.mjs
- **版本与元数据管理**：importance `medium`
  - source_paths: README.md, package.json, landing/docs/index.html, CHANGELOG.md

## Repo Inspection Evidence / 源码检查证据

- repo_clone_verified: true
- repo_inspection_verified: true
- repo_commit: `c8229f458658cc2b21755df88d704e3152f04be9`
- inspected_files: `Dockerfile`, `package.json`, `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 `meridian-mcp` 与安装入口 `meridian-skills-mcp` 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 | mcp_registry:io.github.LuuOW/meridian-skills:1.0.0 | https://registry.modelcontextprotocol.io/v0.1/servers/io.github.LuuOW%2Fmeridian-skills/versions/1.0.0 | repo=meridian-mcp; install=meridian-skills-mcp
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 2: 来源证据：Automate metadata consistency (version, feature-vector dims, URLs)

- Trigger: GitHub 社区证据显示该项目存在一个配置相关的待验证问题：Automate metadata consistency (version, feature-vector dims, URLs)
- Host AI rule: 来源问题仍为 open，Pack Agent 需要复核是否仍影响当前版本。
- Why it matters: 可能增加新用户试用和生产接入成本。
- Evidence: community_evidence:github | cevd_00bb9cc4973344808616920b3eb5c648 | https://github.com/LuuOW/meridian-mcp/issues/907 | 来源类型 github_issue 暴露的待验证使用条件。
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 3: 能力判断依赖假设

- Trigger: README/documentation is current enough for a first validation pass.
- Host AI rule: 将假设转成下游验证清单。
- Why it matters: 假设不成立时，用户拿不到承诺的能力。
- Evidence: capability.assumptions | mcp_registry:io.github.LuuOW/meridian-skills:1.0.0 | https://registry.modelcontextprotocol.io/v0.1/servers/io.github.LuuOW%2Fmeridian-skills/versions/1.0.0 | README/documentation is current enough for a first validation pass.
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 4: 维护活跃度未知

- Trigger: 未记录 last_activity_observed。
- Host AI rule: 补 GitHub 最近 commit、release、issue/PR 响应信号。
- Why it matters: 新项目、停更项目和活跃项目会被混在一起，推荐信任度下降。
- Evidence: evidence.maintainer_signals | mcp_registry:io.github.LuuOW/meridian-skills:1.0.0 | https://registry.modelcontextprotocol.io/v0.1/servers/io.github.LuuOW%2Fmeridian-skills/versions/1.0.0 | last_activity_observed missing
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 5: 下游验证发现风险项

- Trigger: no_demo
- Host AI rule: 进入安全/权限治理复核队列。
- Why it matters: 下游已经要求复核，不能在页面中弱化。
- Evidence: downstream_validation.risk_items | mcp_registry:io.github.LuuOW/meridian-skills:1.0.0 | https://registry.modelcontextprotocol.io/v0.1/servers/io.github.LuuOW%2Fmeridian-skills/versions/1.0.0 | no_demo; severity=medium
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 6: 存在评分风险

- Trigger: no_demo
- Host AI rule: 把风险写入边界卡，并确认是否需要人工复核。
- Why it matters: 风险会影响是否适合普通用户安装。
- Evidence: risks.scoring_risks | mcp_registry:io.github.LuuOW/meridian-skills:1.0.0 | https://registry.modelcontextprotocol.io/v0.1/servers/io.github.LuuOW%2Fmeridian-skills/versions/1.0.0 | 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 | mcp_registry:io.github.LuuOW/meridian-skills:1.0.0 | https://registry.modelcontextprotocol.io/v0.1/servers/io.github.LuuOW%2Fmeridian-skills/versions/1.0.0 | 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 | mcp_registry:io.github.LuuOW/meridian-skills:1.0.0 | https://registry.modelcontextprotocol.io/v0.1/servers/io.github.LuuOW%2Fmeridian-skills/versions/1.0.0 | 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。