# create-vkm-kit - 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 create-vkm-kit 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_0003` 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：`packages/create-vkm-kit/templates/skills/vkm-design/SKILL.md`, `packages/create-vkm-kit/templates/skills/vkm-discipline/SKILL.md`, `packages/create-vkm-kit/templates/skills/vkm-research/SKILL.md`, `packages/create-vkm-kit/templates/skills/vkm-spec/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：`packages/create-vkm-kit/templates/skills/vkm-design/SKILL.md`, `packages/create-vkm-kit/templates/skills/vkm-discipline/SKILL.md`, `packages/create-vkm-kit/templates/skills/vkm-research/SKILL.md`, `packages/create-vkm-kit/templates/skills/vkm-spec/SKILL.md` Claim：`clm_0001` supported 0.86
- **命令行启动或安装流程**（需要安装后验证）：项目文档中存在可执行命令，真实使用需要在本地或宿主环境中运行这些命令。 证据：`AGENTS.md`, `README.md`, `docs/en/install.md`, `packages/create-vkm-kit/README.md` 等 Claim：`clm_0002` supported 0.86

## 怎么开始

- `npx @vkmikc/create-vkm-kit                 # asistente interactivo (pre-marca Codex + Claude)` 证据：`README.md` Claim：`clm_0005` supported 0.86
- `npx @vkmikc/create-vkm-kit -y              # sin preguntas → ~/Documents/obsidian-memory-vault` 证据：`README.md` Claim：`clm_0006` supported 0.86
- `npx @vkmikc/create-vkm-kit "<RUTA>" -y     # sin preguntas, en la ruta que elijas` 证据：`README.md` Claim：`clm_0007` supported 0.86
- `npx @vkmikc/create-vkm-kit@latest -y` 证据：`packages/create-vkm-kit/README.md` Claim：`clm_0008` supported 0.86
- `npx @vkmikc/create-vkm-kit@latest ./my-vault -y` 证据：`packages/create-vkm-kit/README.md` Claim：`clm_0009` supported 0.86, `clm_0010` supported 0.86, `clm_0011` supported 0.86
- `npx @vkmikc/create-vkm-kit@latest ./my-vault -y --minimal` 证据：`packages/create-vkm-kit/README.md` Claim：`clm_0010` supported 0.86
- `npx @vkmikc/create-vkm-kit@latest ./my-vault -y --ide cursor,claude --rules all` 证据：`packages/create-vkm-kit/README.md` Claim：`clm_0011` supported 0.86
- `git clone --depth 1 https://github.com/searxng/searxng.git $HOME\.vkm\searxng-src` 证据：`packages/obscura-web/searxng/README.md` Claim：`clm_0012` supported 0.86
- `npx @vkmikc/create-vkm-kit --obscura        # add the obscura web MCP` 证据：`packages/obscura-web/README.md` Claim：`clm_0013` supported 0.86
- `npx @vkmikc/create-vkm-kit --full           # everything, obscura included` 证据：`packages/obscura-web/README.md` Claim：`clm_0014` supported 0.86

## 继续前判断卡

- **当前建议**：先做研究框架试用
- **为什么**：这个项目面向研究工作流，核心风险是资料可信度和输出质量；先用 Prompt Preview 验证研究框架，再在隔离环境试装。

### 30 秒判断

- **现在怎么做**：先做研究框架试用
- **最小安全下一步**：先用 Prompt Preview 验证研究框架；满意后再隔离试装
- **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_0003` 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：`packages/create-vkm-kit/templates/skills/vkm-design/SKILL.md`, `packages/create-vkm-kit/templates/skills/vkm-discipline/SKILL.md`, `packages/create-vkm-kit/templates/skills/vkm-research/SKILL.md`, `packages/create-vkm-kit/templates/skills/vkm-spec/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：`packages/create-vkm-kit/templates/skills/vkm-design/SKILL.md`, `packages/create-vkm-kit/templates/skills/vkm-discipline/SKILL.md`, `packages/create-vkm-kit/templates/skills/vkm-research/SKILL.md`, `packages/create-vkm-kit/templates/skills/vkm-spec/SKILL.md` Claim：`clm_0001` supported 0.86
- **能力存在：命令行启动或安装流程**（supported）：可以相信项目包含这类能力线索；是否适合你的具体任务仍要试用或安装后验证。 证据：`AGENTS.md`, `README.md`, `docs/en/install.md`, `packages/create-vkm-kit/README.md` 等 Claim：`clm_0002` supported 0.86
- **存在 Quick Start / 安装命令线索**（supported）：可以相信项目文档出现过启动或安装入口；不要因此直接在主力环境运行。 证据：`README.md` Claim：`clm_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：`AGENTS.md`, `packages/create-vkm-kit/templates/skills/vkm-design/SKILL.md`, `packages/create-vkm-kit/templates/skills/vkm-discipline/SKILL.md`, `packages/create-vkm-kit/templates/skills/vkm-research/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：`AGENTS.md`, `README.md`, `docs/en/install.md`, `packages/create-vkm-kit/README.md` 等
- **宿主 AI 配置**：Claude/Codex/Cursor/Gemini/OpenCode 等宿主的 plugin、Skill 或规则加载配置。 原因：宿主配置会改变 AI 后续工作方式，可能和用户已有规则冲突。 证据：`AGENTS.md`, `packages/create-vkm-kit/templates/skills/vkm-design/SKILL.md`, `packages/create-vkm-kit/templates/skills/vkm-discipline/SKILL.md`, `packages/create-vkm-kit/templates/skills/vkm-research/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：`AGENTS.md`, `README.md`, `docs/en/install.md`, `packages/create-vkm-kit/README.md` 等
- **宿主 AI 上下文**：AI Context Pack、Prompt Preview、Skill 路由、风险规则和项目事实。 原因：导入上下文会影响宿主 AI 后续判断，必须避免把未验证项包装成事实。

### 最小安全下一步

- **先跑 Prompt Preview**：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。）
- **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。
- **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_0024` 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：`AGENTS.md`, `README.md`, `docs/en/install.md`, `packages/create-vkm-kit/README.md` 等 Claim：`clm_0025` 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：`packages/create-vkm-kit/templates/skills/vkm-design/SKILL.md`, `packages/create-vkm-kit/templates/skills/vkm-discipline/SKILL.md`, `packages/create-vkm-kit/templates/skills/vkm-research/SKILL.md`, `packages/create-vkm-kit/templates/skills/vkm-spec/SKILL.md` Claim：`clm_0001` supported 0.86
- **命令行启动或安装流程**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`AGENTS.md`, `README.md`, `docs/en/install.md`, `packages/create-vkm-kit/README.md` 等 Claim：`clm_0002` supported 0.86

### 上下文规模

- 文件总数：526
- 重要文件覆盖：40/526
- 证据索引条目：80
- 角色 / Skill 条目：4

### 证据不足时的处理

- **missing_evidence**：说明证据不足，要求用户提供目标文件、README 段落或安装后验证记录；不要补全事实。
- **out_of_scope_request**：说明该任务超出当前 AI Context Pack 证据范围，并建议用户先查看 Human Manual 或真实安装后验证。
- **runtime_request**：给出安装前检查清单和命令来源，但不要替用户执行命令或声称已执行。
- **source_conflict**：同时展示冲突来源，标记为待核实，不要强行选择一个版本。

## Prompt Recipes

### 适配判断

- 目标：判断这个项目是否适合用户当前任务。
- 预期输出：适配结论、关键理由、证据引用、安装前可预览内容、必须安装后验证内容、下一步建议。

```text
请基于 create-vkm-kit 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 create-vkm-kit 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 create-vkm-kit 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。

- **vkm-design**（skill）：Professional, anti-generic UI/UX and visual design method for any medium — web, desktop, dataviz, brand. Forces a named design direction before code, computed checks contrast, scale , real current libraries, and a visual self-critique loop. Invoke for any design, styling, or UI task. 激活提示：当用户任务与“vkm-design”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：`packages/create-vkm-kit/templates/skills/vkm-design/SKILL.md`
- **vkm-discipline**（skill）：Cross-domain execution discipline — infer the real intent, do it the best way, and hand back more than the literal ask, with minimal friction; depth scaled to task difficulty and model. Bias to action; ask only when the answer changes what you'd do. Invoke on any non-trivial task. 激活提示：当用户任务与“vkm-discipline”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：`packages/create-vkm-kit/templates/skills/vkm-discipline/SKILL.md`
- **vkm-research**（skill）：Consolidate a RESEARCH/ bank into a quality summary.md — read the hub + unconsolidated sources, synthesize with wikilinks and supersedes, mark what's done. Also the import path for external research reports. Invoke for "consolidate " or "write up the research on X". 激活提示：当用户任务与“vkm-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：`packages/create-vkm-kit/templates/skills/vkm-research/SKILL.md`
- **vkm-spec**（skill）：Turn a vague idea into a precise, context-grounded spec before implementing — one assemble context call + the orchestration template, filled in-session. 激活提示：当用户任务与“vkm-spec”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：`packages/create-vkm-kit/templates/skills/vkm-spec/SKILL.md`

## 证据索引

- 共索引 80 条证据。

- **Documentación · Documentation**（documentation）：Elige tu idioma · Pick your language: 证据：`docs/README.md`
- **Architecture Decision Records**（documentation）：Each file in this directory captures one design decision: what was chosen, what alternatives existed, and why the chosen path won. Early ADRs reflect the original Windows-first kit; ADR-0010 onward document the current cross-platform stack. 证据：`docs/adr/README.md`
- **Documentation English**（documentation）：🇪🇸 Español ../es/README.md · 🇬🇧 English 证据：`docs/en/README.md`
- **Documentación español**（documentation）：🇪🇸 Español · 🇬🇧 English ../en/README.md 证据：`docs/es/README.md`
- **Legacy / Archivo histórico**（documentation）：🇪🇸 Estas notas describen versiones anteriores del kit. No las necesitas para instalar la versión actual. Se conservan para quien migra desde v1/v2 y para trazar por qué el diseño es como es hoy. 🇬🇧 These notes describe older versions of the kit. You don't need them to install the current version. They're kept for anyone migrating from v1/v2 and to trace why today's design is what it is. 证据：`docs/legacy/README.md`
- **AGENTS.md**（documentation）：Canonical instructions for any AI agent or IDE consuming this repository Cursor, Claude Code, Copilot, Codex CLI, Aider, Windsurf, Zed, Continue, etc. . Humans onboarding this kit: read docs/en/install.md ./docs/en/install.md linear then docs/en/how-it-works.md ./docs/en/how-it-works.md . For human onboarding of the pattern vault + MCP , read README.md . 证据：`AGENTS.md`
- **¿Qué es esto? · What is this?**（documentation）：🧠 Memoria persistente para tu agente de IA Persistent memory for your AI agent 证据：`README.md`
- **Evaluations**（documentation）：Three distinct things live here. Keep them separate: 证据：`evals/README.md`
- **Examples**（documentation）：A walkthrough of what a healthy vault looks like a few weeks into use. All names and projects are fictional. 证据：`examples/README.md`
- **scripts/**（documentation）：- sync-agents.ts — regenerates AGENTS.md , .cursor/rules , .continue/rules from .agents/ and agents-manifest.yaml . - linkcheck.mjs — offline relative-link + heading-anchor checker for the Markdown docs npm run linkcheck . Complements the lychee CI job, which also validates external URLs. - license-sync.mjs — mirrors the root LICENSE.md into every packages/ dir npm run license:sync ; --check gates CI and npm publish . npm only auto-includes LICENSE files from the package root, so the published tarball needs its own copy. 证据：`scripts/README.md`
- **tools/**（documentation）：Windows helper .ps1 scripts that used to live here were removed in v3 see docs/legacy/v2-to-v3-script-free-kit.md ../docs/legacy/v2-to-v3-script-free-kit.md . Use the troubleshooting guide and Task Manager / Resource Monitor for the same outcomes. 证据：`tools/README.md`
- **design-bench — does the vkm-design skill actually improve output?**（documentation）：design-bench — does the vkm-design skill actually improve output? 证据：`evals/design-bench/README.md`
- **/vkm-discipline mini-bench**（documentation）：Measures whether loading the /vkm-discipline doctrine helps or hurts task quality versus stock, per model — the parity check before the kit's discipline system supersedes the SOP-suite. The key question, given that heavy procedure/CoT measurably hurts small models: does +discipline ≥ stock on the small model Haiku , while still helping the large one Opus ? 证据：`evals/discipline-bench/README.md`
- **@vkmikc/create-vkm-kit**（documentation）：Interactive initializer for Obsidian-style, file-based agent memory — a Markdown vault your AI coding agent reads and writes across sessions, wired to your IDE over MCP . 证据：`packages/create-vkm-kit/README.md`
- **@vkmikc/obscura-web**（documentation）：Stealth web access for your AI agent — an opt-in MCP server that fetches and searches the web through the local obscura https://github.com/h4ckf0r0day/obscura headless browser, preferred over the native WebFetch / WebSearch soft enforcement — the native tools stay as the fallback . 证据：`packages/obscura-web/README.md`
- **Local SearXNG for obscura search structured, fast, relevant search**（documentation）：Local SearXNG for obscura search structured, fast, relevant search 证据：`packages/obscura-web/searxng/README.md`
- **obsidian-memory-mcp**（documentation）：The hybrid MCP sidecar Node, stdio : 15 vault tools for any MCP-capable agent — vault hybrid search BM25 + semantic via RRF, passage-first, compact wire — ADR-0034 , assemble context single-call budgeted context — ADR-0045 , vault fts search , vault complete , the typed knowledge graph vault relations / vault observations / vault kg suggest , vault-locked file tools, vault audit , vault memory report and memory extract candidates . Search/index/KG calls bridge to the Python engine obsidian-memory-rag ../obsidian-memory-rag/ . 证据：`packages/obsidian-memory-mcp/README.md`
- **obsidian-memory-rag**（documentation）：Optional, dependency-free retrieval sidecar for Markdown vaults: SQLite FTS5 / BM25 lexical search ADR-0014 , hybrid BM25 + semantic-vector ranking fused with Reciprocal Rank Fusion ADR-0017 , and graph-aware recall over your wikilinks ADR-0019 . Complements basic-memory with incremental indexing and passage-first results it returns the matching section , not the whole note — the main token saver . 证据：`packages/obsidian-memory-rag/README.md`
- **@vkmikc/vkm-doctor**（documentation）：Local token & cache doctor for Claude Code — part of the vkm-kit https://github.com/Vahlame/create-vkm-kit efficiency suite ADR-0044 . Everything stays on your machine: no cloud collector, no accounts. 证据：`packages/vkm-doctor/README.md`
- **@vkmikc/vkm-downloads**（documentation）：A guarded file-download manager for your AI agent — an opt-in MCP server that lets the agent download files from the web to ~/Downloads/vkm-kit/ , with the guardrails that make that safe. 证据：`packages/vkm-downloads/README.md`
- **@vkmikc/vkm-spec**（documentation）：Idea → spec builder of the vkm-kit https://github.com/Vahlame/create-vkm-kit suite: turns a one-line idea into a precise, vault-grounded XML you paste into Claude Code or any agent — so execution starts from a real spec instead of a vague prompt ADR-0046 . 证据：`packages/vkm-spec/README.md`
- **Package**（package_manifest）：{ "name": "vkm-kit", "private": true, "workspaces": "packages/ " , "scripts": { "setup": "node scripts/agent-install.mjs", "setup:dry": "node scripts/agent-install.mjs --dry-run", "sync-agents": "npx --yes tsx scripts/sync-agents.ts", "sync-agents:check": "npx --yes tsx scripts/sync-agents.ts --check", "lint": "eslint .", "typecheck": "tsc -p tsconfig.json && tsc -p tsconfig.checkjs.json", "license:sync": "node scripts/license-sync.mjs", "license:sync:check": "node scripts/license-sync.mjs --check", "linkcheck": "node scripts/linkcheck.mjs .", "preview:assets": "node scripts/preview-assets.mjs", "eval:adherence": "node evals/run-adherence-ci.mjs", "version:check": "node scripts/version.mjs… 证据：`package.json`
- **Package**（package_manifest）：{ "name": "@vkmikc/create-obsidian-memory", "version": "4.4.0", "description": "RENAMED: this package is now @vkmikc/create-vkm-kit the vkm-kit installer . This shim forwards every invocation there — update your scripts.", "type": "module", "license": "SEE LICENSE IN LICENSE.md", "engines": { "node": " =20" }, "bin": { "create-obsidian-memory": "src/index.js" }, "files": "src" , "scripts": { "test": "node --test" }, "publishConfig": { "access": "public" }, "dependencies": { "execa": "^10.0.0", "picocolors": "^1.1.1" } } 证据：`packages/create-obsidian-memory-shim/package.json`
- **Package**（package_manifest）：{ "name": "@vkmikc/create-vkm-kit", "version": "4.4.0", "description": "vkm-kit installer: persistent Markdown+git memory for AI agents MCP with hybrid search, stealth web fetch/search obscura + on-demand SearXNG , a cross-domain execution-discipline skill, token-saver hooks, a local usage doctor and a spec-builder — one npx command.", "license": "SEE LICENSE IN LICENSE.md", "homepage": "https://github.com/Vahlame/create-vkm-kit readme", "repository": { "type": "git", "url": "git+https://github.com/Vahlame/create-vkm-kit.git", "directory": "packages/create-vkm-kit" }, "keywords": "mcp", "obsidian", "cursor", "claude-code", "codex", "agent-memory", "ai-memory", "basic-memory", "rag", "semant… 证据：`packages/create-vkm-kit/package.json`
- **Package**（package_manifest）：{ "name": "vkm-design-illustration-deps", "private": true, "description": "Pinned deps for trace-svg.mjs and treat-photo.mjs. jimp MUST stay on the 0.22.x line: jimp v1 current npm default renamed/removed the exact API these scripts use default export, positional constructor, Jimp.MIME PNG, Jimp.AUTO, getBufferAsync . potrace itself depends on jimp@^0.14.0, so it never touches v1 either. Run npm install in this folder before using either script; the rest of the skill needs zero npm deps.", "dependencies": { "jimp": "0.22.x", "potrace": "2.1.8", "svg-path-bbox": "2.1.0" } } 证据：`packages/create-vkm-kit/templates/skills/vkm-design/scripts/package.json`
- **Package**（package_manifest）：{ "name": "@vkmikc/obscura-web", "version": "4.4.0", "private": true, "license": "SEE LICENSE IN LICENSE.md", "type": "module", "bin": { "obscura-web-mcp": "./src/obscura-mcp.mjs" }, "scripts": { "test": "node --test", "start": "node ./src/obscura-mcp.mjs" }, "dependencies": { "@modelcontextprotocol/sdk": "^1.29.0", "cheerio": "^1.2.0", "execa": "^10.0.0", "zod": "^4.4.3" } } 证据：`packages/obscura-web/package.json`
- **Package**（package_manifest）：{ "name": "@vkmikc/obsidian-memory-mcp", "version": "4.4.0", "private": true, "license": "SEE LICENSE IN LICENSE.md", "type": "module", "bin": { "obsidian-memory-hybrid-mcp": "./src/hybrid-mcp.mjs" }, "scripts": { "test": "node --test", "start:hybrid": "node ./src/hybrid-mcp.mjs" }, "dependencies": { "@modelcontextprotocol/sdk": "^1.29.0", "execa": "^10.0.0", "pino": "^10.3.1", "zod": "^4.4.3" }, "optionalDependencies": { "@opentelemetry/api": "^1.9.0", "@opentelemetry/exporter-trace-otlp-http": "^0.219.0", "@opentelemetry/sdk-node": "^0.219.0" } } 证据：`packages/obsidian-memory-mcp/package.json`
- **Package**（package_manifest）：{ "name": "@vkmikc/vkm-doctor", "version": "4.4.0", "description": "Local token/cache doctor for Claude Code: OTLP/HTTP JSON sink localhost-only, NDJSON storage + usage report CLI. Part of the vkm-kit efficiency suite.", "type": "module", "private": true, "license": "SEE LICENSE IN LICENSE.md", "engines": { "node": " =20" }, "bin": { "vkm-doctor": "src/doctor.mjs", "vkm-otel-sink": "src/otel-sink.mjs" }, "files": "src" , "scripts": { "test": "node --test" }, "dependencies": { "@vkmikc/create-vkm-kit": " " } } 证据：`packages/vkm-doctor/package.json`
- **Package**（package_manifest）：{ "name": "@vkmikc/vkm-downloads", "version": "4.4.0", "private": true, "license": "SEE LICENSE IN LICENSE.md", "type": "module", "bin": { "vkm-downloads-mcp": "./src/downloads-mcp.mjs" }, "scripts": { "test": "node --test", "start": "node ./src/downloads-mcp.mjs" }, "dependencies": { "@modelcontextprotocol/sdk": "^1.29.0", "zod": "^4.4.3" } } 证据：`packages/vkm-downloads/package.json`
- **Package**（package_manifest）：{ "name": "@vkmikc/vkm-spec", "version": "4.4.0", "private": true, "license": "SEE LICENSE IN LICENSE.md", "description": "Spec-builder pipeline of the vkm-kit suite: idea → vault-aware XML, with an optional local Ollama phi4-mini draft pass and a deterministic fallback that always works when Ollama is absent ADR-0046/0047/0048 .", "type": "module", "bin": { "vkm-spec": "src/cli.mjs" }, "files": "src", "public" , "scripts": { "test": "node --test", "gui": "node src/server.mjs" }, "dependencies": { "@vkmikc/obsidian-memory-mcp": " ", "clipboardy": "^5.0.0", "picocolors": "^1.1.1", "prompts": "^2.4.2", "zod": "^4.4.3" }, "engines": { "node": " =20" } } 证据：`packages/vkm-spec/package.json`
- **Installation step by step, 100% repeatable**（documentation）：🇪🇸 Español ../es/instalacion.md · 🇬🇧 English 证据：`docs/en/install.md`
- **vkm-design — design like a trained professional**（skill_instruction）：vkm-design — design like a trained professional 证据：`packages/create-vkm-kit/templates/skills/vkm-design/SKILL.md`
- **vkm-discipline — resourceful execution**（skill_instruction）：vkm-discipline — resourceful execution 证据：`packages/create-vkm-kit/templates/skills/vkm-discipline/SKILL.md`
- **vkm-research — consolidate a RESEARCH/ topic into a quality summary**（skill_instruction）：vkm-research — consolidate a RESEARCH/ topic into a quality summary 证据：`packages/create-vkm-kit/templates/skills/vkm-research/SKILL.md`
- **/vkm-spec — sharpen an idea into a spec, in-session**（skill_instruction）：/vkm-spec — sharpen an idea into a spec, in-session 证据：`packages/create-vkm-kit/templates/skills/vkm-spec/SKILL.md`
- **ADR-0053: vkm-design — a model-agnostic professional design skill**（skill_instruction）：ADR-0053: vkm-design — a model-agnostic professional design skill 证据：`docs/adr/0053-vkm-design-skill.md`
- **Contributing**（documentation）：Thanks for considering a contribution. This repository ships cross-platform guidance AGENTS.md and optional Go and Node tooling. 证据：`CONTRIBUTING.md`
- **Architecture**（documentation）：This document is the map of the repository: what the pieces are, how data flows between them, and the design patterns they share. The rationale for each decision lives in docs/adr/ ./docs/adr/ — when this document says "because X", the ADR is where "X" is argued in full. For agent-facing operating instructions see AGENTS.md ./AGENTS.md ; for human onboarding see docs/en/install.md ./docs/en/install.md . 证据：`ARCHITECTURE.md`
- **What is this?**（documentation）：🧠 Persistent memory for your AI agent Memoria persistente para tu agente de IA 证据：`README.en.md`
- **Observability**（documentation）：The kit has three observability surfaces. All of them are local — nothing leaves your machine. 证据：`docs/observability.md`
- **ADR-0001: Use mcp-remote instead of pointing Cursor at the SSE server directly**（documentation）：ADR-0001: Use mcp-remote instead of pointing Cursor at the SSE server directly 证据：`docs/adr/0001-use-mcp-remote-instead-of-direct-sse.md`
- **ADR-0002: Run the MCP server as a separate process**（documentation）：ADR-0002: Run the MCP server as a separate process 证据：`docs/adr/0002-run-mcp-server-as-separate-process.md`
- **ADR-0003: Run scheduled tasks via wscript.exe and a VBS shim**（documentation）：ADR-0003: Run scheduled tasks via wscript.exe and a VBS shim 证据：`docs/adr/0003-scheduled-tasks-via-wscript.md`
- **ADR-0004: Sync order is add - commit - pull --rebase - push**（documentation）：ADR-0004: Sync order is add - commit - pull --rebase - push 证据：`docs/adr/0004-sync-order-add-commit-pull-push.md`
- **ADR-0005: Use PowerShell 5.1-compatible JSON merging**（documentation）：ADR-0005: Use PowerShell 5.1-compatible JSON merging 证据：`docs/adr/0005-powershell-5-compatible-json-merge.md`
- **ADR-0006: Scripts live in the user's vault, not in this repo**（documentation）：ADR-0006: Scripts live in the user's vault, not in this repo 证据：`docs/adr/0006-no-runnable-scripts-in-this-repo.md`
- **ADR-0007: Windows-first; other platforms via separate prompt variants**（documentation）：ADR-0007: Windows-first; other platforms via separate prompt variants 证据：`docs/adr/0007-windows-first-pattern.md`
- **ADR-0008: Ship Vault-Doctor.ps1 as the canonical vault health audit**（documentation）：ADR-0008: Ship Vault-Doctor.ps1 as the canonical vault health audit 证据：`docs/adr/0008-vault-doctor-as-canonical-tool.md`
- **ADR-0009: Default vault layout with YAML frontmatter and a three-level reading flow**（documentation）：ADR-0009: Default vault layout with YAML frontmatter and a three-level reading flow 证据：`docs/adr/0009-frontmatter-and-three-level-reading-flow.md`
- **ADR-0010: Migrate MCP stack to basic-memory Streamable HTTP**（documentation）：ADR-0010: Migrate MCP stack to basic-memory Streamable HTTP 证据：`docs/adr/0010-migrate-to-basic-memory.md`
- **ADR-0011: AGENTS.md as canonical agent surface IDE-agnostic**（documentation）：ADR-0011: AGENTS.md as canonical agent surface IDE-agnostic 证据：`docs/adr/0011-adopt-agents-md.md`
- **ADR-0012: Cross-platform Go daemon obsidian-memoryd replaces PowerShell + Task Scheduler**（documentation）：ADR-0012: Cross-platform Go daemon obsidian-memoryd replaces PowerShell + Task Scheduler 证据：`docs/adr/0012-go-daemon-cross-platform.md`
- **ADR-0013: Syncthing as an optional sync transport**（documentation）：ADR-0013: Syncthing as an optional sync transport 证据：`docs/adr/0013-syncthing-as-transport.md`
- **ADR-0014: Hybrid retrieval FTS5 + sqlite-vec as optional RAG sidecar**（documentation）：ADR-0014: Hybrid retrieval FTS5 + sqlite-vec as optional RAG sidecar 证据：`docs/adr/0014-hybrid-retrieval-sqlite-vec.md`
- **ADR-0015: Keep privacy guidance generic in public docs**（documentation）：ADR-0015: Keep privacy guidance generic in public docs 证据：`docs/adr/0015-privacy-compliance-documentation.md`
- **ADR-0016: Default localhost port for Streamable HTTP basic-memory**（documentation）：ADR-0016: Default localhost port for Streamable HTTP basic-memory 证据：`docs/adr/0016-localhost-mcp-default-port.md`
- **ADR-0017: Hybrid query — pluggable embeddings + pure-Python cosine**（documentation）：ADR-0017: Hybrid query — pluggable embeddings + pure-Python cosine 证据：`docs/adr/0017-hybrid-query-embeddings.md`
- **ADR-0018: Multi-agent token efficiency — passage-first reads, auto-indexed search, untrusted-data envelope**（documentation）：ADR-0018: Multi-agent token efficiency — passage-first reads, auto-indexed search, untrusted-data envelope 证据：`docs/adr/0018-multi-agent-token-efficiency.md`
- **ADR-0019: Graph-aware retrieval over the wikilink graph**（documentation）：ADR-0019: Graph-aware retrieval over the wikilink graph 证据：`docs/adr/0019-graph-aware-retrieval.md`
- **ADR-0020: Measured retrieval quality recall@k / MRR as a CI gate**（documentation）：ADR-0020: Measured retrieval quality recall@k / MRR as a CI gate 证据：`docs/adr/0020-measured-retrieval-quality.md`
- 其余 20 条证据见 `AI_CONTEXT_PACK.json` 或 `EVIDENCE_INDEX.json`。

## 宿主 AI 必须遵守的规则

- **把本资产当作开工前上下文，而不是运行环境。**：AI Context Pack 只包含证据化项目理解，不包含目标项目的可执行状态。 证据：`docs/README.md`, `docs/adr/README.md`, `docs/en/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：`docs/README.md`, `docs/adr/README.md`, `docs/en/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。

- **项目概述与系统架构**：importance `high`
  - source_paths: README.md, README.en.md, ARCHITECTURE.md, AGENTS.md, package.json
- **持久化记忆、混合检索与多写者安全**：importance `high`
  - source_paths: packages/obsidian-memory-mcp/src/hybrid-mcp.mjs, packages/obsidian-memory-mcp/src/vault-fs.mjs, packages/obsidian-memory-mcp/src/vault-lock.mjs, packages/obsidian-memory-mcp/src/context-assemble.mjs, packages/obsidian-memory-mcp/src/rag-client.mjs
- **Token 节约、本地诊断与 CI 守护**：importance `high`
  - source_paths: packages/create-vkm-kit/src/token-saver.mjs, packages/create-vkm-kit/src/hooks/compact-tool-output.mjs, packages/create-vkm-kit/src/hooks/compact-mcp-output.mjs, packages/create-vkm-kit/src/hooks/ensure-otel-sink.mjs, packages/create-vkm-kit/src/hooks/guard-native-memory-write.mjs
- **技能体系、Spec 与外部集成**：importance `high`
  - source_paths: packages/create-vkm-kit/src/skills-install.mjs, packages/create-vkm-kit/templates/skills/vkm-discipline/SKILL.md, packages/create-vkm-kit/templates/skills/vkm-spec/SKILL.md, packages/create-vkm-kit/templates/skills/vkm-design/SKILL.md, packages/create-vkm-kit/templates/skills/vkm-research/SKILL.md

## Repo Inspection Evidence / 源码检查证据

- repo_clone_verified: true
- repo_inspection_verified: true
- repo_commit: `9fbefd63cf22de857dd618c9314ecdb058603951`
- inspected_files: `README.md`, `package.json`, `docs/README.md`, `docs/adr/0001-use-mcp-remote-instead-of-direct-sse.md`, `docs/adr/0002-run-mcp-server-as-separate-process.md`, `docs/adr/0003-scheduled-tasks-via-wscript.md`, `docs/adr/0004-sync-order-add-commit-pull-push.md`, `docs/adr/0005-powershell-5-compatible-json-merge.md`, `docs/adr/0006-no-runnable-scripts-in-this-repo.md`, `docs/adr/0007-windows-first-pattern.md`, `docs/adr/0008-vault-doctor-as-canonical-tool.md`, `docs/adr/0009-frontmatter-and-three-level-reading-flow.md`, `docs/adr/0010-migrate-to-basic-memory.md`, `docs/adr/0011-adopt-agents-md.md`, `docs/adr/0012-go-daemon-cross-platform.md`, `docs/adr/0013-syncthing-as-transport.md`, `docs/adr/0014-hybrid-retrieval-sqlite-vec.md`, `docs/adr/0015-privacy-compliance-documentation.md`, `docs/adr/0016-localhost-mcp-default-port.md`, `docs/adr/0017-hybrid-query-embeddings.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: 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/Vahlame/create-vkm-kit | 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 2: 能力判断依赖假设

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

### Constraint 3: 维护活跃度未知

- Trigger: 未记录 last_activity_observed。
- Host AI rule: 补 GitHub 最近 commit、release、issue/PR 响应信号。
- Why it matters: 新项目、停更项目和活跃项目会被混在一起，推荐信任度下降。
- Evidence: evidence.maintainer_signals | https://github.com/Vahlame/create-vkm-kit | 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。

- Trigger: no_demo
- Evidence: downstream_validation.risk_items | https://github.com/Vahlame/create-vkm-kit | no_demo; severity=medium
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 5: 存在评分风险

- Trigger: no_demo
- Why it matters: 风险会影响是否适合普通用户安装。
- Evidence: risks.scoring_risks | https://github.com/Vahlame/create-vkm-kit | 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: 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 | https://github.com/Vahlame/create-vkm-kit | 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 7: 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 | https://github.com/Vahlame/create-vkm-kit | 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。