# nano-brain - 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 nano-brain 编译的 AI Context Pack。请把它当作开工前上下文：帮助用户理解适合谁、能做什么、如何开始、哪些必须安装后验证、风险在哪里。不要声称你已经安装、运行或执行了目标项目。

## Claim 消费规则

- **事实来源**：Repo Evidence + Claim/Evidence Graph；Human Wiki 只提供显著性、术语和叙事结构。
- **事实最低状态**：`supported`
- `supported`：可以作为项目事实使用，但回答中必须引用 claim_id 和证据路径。
- `weak`：只能作为低置信度线索，必须要求用户继续核实。
- `inferred`：只能用于风险提示或待确认问题，不能包装成项目事实。
- `unverified`：不得作为事实使用，应明确说证据不足。
- `contradicted`：必须展示冲突来源，不得替用户强行选择一个版本。

## 它最适合谁

- **希望把专业流程带进宿主 AI 的用户**：仓库包含 Skill 文档。 证据：`.claude/skills/openspec-apply-change/SKILL.md`, `.claude/skills/openspec-archive-change/SKILL.md`, `.claude/skills/openspec-explore/SKILL.md`, `.claude/skills/openspec-propose/SKILL.md` 等 Claim：`clm_0003` supported 0.86

## 它能做什么

- **AI Skill / Agent 指令资产库**（可做安装前预览）：项目包含可被宿主 AI 读取的 Skill 或 Agent 指令文件，可用于把专业流程带入 Claude、Codex、Cursor 等宿主。 证据：`.claude/skills/openspec-apply-change/SKILL.md`, `.claude/skills/openspec-archive-change/SKILL.md`, `.claude/skills/openspec-explore/SKILL.md`, `.claude/skills/openspec-propose/SKILL.md` 等 Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86
- **命令行启动或安装流程**（需要安装后验证）：项目文档中存在可执行命令，真实使用需要在本地或宿主环境中运行这些命令。 证据：`.opencode/skills/harness-init/SKILL.md`, `.opencode/skills/nano-brain/references/http-api.md`, `.planning/phases/08-session-harvest-unification-ticket-linking/08-03-PLAN.md`, `.planning/phases/13-interactive-init-wizard-one-command-interactive-setup-detect/13-08-PLAN.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86

## 怎么开始

- `npm install -g @openspec/cli` 证据：`.opencode/skills/harness-init/SKILL.md` Claim：`clm_0004` supported 0.86
- `curl -sf "$BASE/health" >/dev/null && echo "ready"` 证据：`.opencode/skills/nano-brain/references/http-api.md` Claim：`clm_0005` unverified 0.25
- `curl -s "$BASE/api/status" | jq` 证据：`.opencode/skills/nano-brain/references/http-api.md` Claim：`clm_0006` unverified 0.25
- `curl -s -X POST "$BASE/api/v1/search" \` 证据：`.opencode/skills/nano-brain/references/http-api.md` Claim：`clm_0007` unverified 0.25
- `curl -s -X POST "$BASE/api/v1/vsearch" \` 证据：`.opencode/skills/nano-brain/references/http-api.md` Claim：`clm_0008` unverified 0.25
- `curl -s -X POST "$BASE/api/v1/query" \` 证据：`.opencode/skills/nano-brain/references/http-api.md` Claim：`clm_0009` unverified 0.25
- `curl -s -X POST "$BASE/api/v1/write" \` 证据：`.opencode/skills/nano-brain/references/http-api.md` Claim：`clm_0010` unverified 0.25
- `curl -s -X POST "$BASE/api/harvest" | jq` 证据：`.opencode/skills/nano-brain/references/http-api.md` Claim：`clm_0011` unverified 0.25
- `curl -s -X POST "$BASE/api/v1/reindex" \` 证据：`.opencode/skills/nano-brain/references/http-api.md` Claim：`clm_0012` unverified 0.25
- `curl -s -X POST "$BASE/api/reload-config" | jq` 证据：`.opencode/skills/nano-brain/references/http-api.md` Claim：`clm_0013` unverified 0.25

## 继续前判断卡

- **当前建议**：仅建议沙盒试装
- **为什么**：项目存在安装命令、宿主配置或本地写入线索，不建议直接进入主力环境，应先在隔离环境试装。

### 30 秒判断

- **现在怎么做**：仅建议沙盒试装
- **最小安全下一步**：先跑 Prompt Preview；若仍要安装，只在隔离环境试装
- **先别相信**：真实输出质量不能在安装前相信。
- **继续会触碰**：命令执行、宿主 AI 配置、本地环境或项目文件

### 现在可以相信

- **适合人群线索：希望把专业流程带进宿主 AI 的用户**（supported）：有 supported claim 或项目证据支撑，但仍不等于真实安装效果。 证据：`.claude/skills/openspec-apply-change/SKILL.md`, `.claude/skills/openspec-archive-change/SKILL.md`, `.claude/skills/openspec-explore/SKILL.md`, `.claude/skills/openspec-propose/SKILL.md` 等 Claim：`clm_0003` supported 0.86
- **能力存在：AI Skill / Agent 指令资产库**（supported）：可以相信项目包含这类能力线索；是否适合你的具体任务仍要试用或安装后验证。 证据：`.claude/skills/openspec-apply-change/SKILL.md`, `.claude/skills/openspec-archive-change/SKILL.md`, `.claude/skills/openspec-explore/SKILL.md`, `.claude/skills/openspec-propose/SKILL.md` 等 Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86
- **能力存在：命令行启动或安装流程**（supported）：可以相信项目包含这类能力线索；是否适合你的具体任务仍要试用或安装后验证。 证据：`.opencode/skills/harness-init/SKILL.md`, `.opencode/skills/nano-brain/references/http-api.md`, `.planning/phases/08-session-harvest-unification-ticket-linking/08-03-PLAN.md`, `.planning/phases/13-interactive-init-wizard-one-command-interactive-setup-detect/13-08-PLAN.md` Claim：`clm_0001` supported 0.86, `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：`.opencode/skills/harness-init/SKILL.md` Claim：`clm_0004` supported 0.86

### 现在还不能相信

- **真实输出质量不能在安装前相信。**（unverified）：Prompt Preview 只能展示引导方式，不能证明真实项目中的结果质量。
- **宿主 AI 版本兼容性不能在安装前相信。**（unverified）：Claude、Cursor、Codex、Gemini 等宿主加载规则和版本差异必须在真实环境验证。
- **不会污染现有宿主 AI 行为，不能直接相信。**（inferred）：Skill、plugin、AGENTS/CLAUDE/GEMINI 指令可能改变宿主 AI 的默认行为。 证据：`.claude/skills/openspec-apply-change/SKILL.md`, `.claude/skills/openspec-archive-change/SKILL.md`, `.claude/skills/openspec-explore/SKILL.md`, `.claude/skills/openspec-propose/SKILL.md` 等
- **可安全回滚不能默认相信。**（unverified）：除非项目明确提供卸载和恢复说明，否则必须先在隔离环境验证。
- **真实安装后是否与用户当前宿主 AI 版本兼容？**（unverified）：兼容性只能通过实际宿主环境验证。
- **项目输出质量是否满足用户具体任务？**（unverified）：安装前预览只能展示流程和边界，不能替代真实评测。
- **安装命令是否需要网络、权限或全局写入？**（unverified）：这影响企业环境和个人环境的安装风险。 证据：`.opencode/skills/harness-init/SKILL.md`

### 继续会触碰什么

- **命令执行**：包管理器、网络下载、本地插件目录、项目配置或用户主目录。 原因：运行第一条命令就可能产生环境改动；必须先判断是否值得跑。 证据：`.opencode/skills/harness-init/SKILL.md`, `.opencode/skills/nano-brain/references/http-api.md`, `.planning/phases/08-session-harvest-unification-ticket-linking/08-03-PLAN.md`, `.planning/phases/13-interactive-init-wizard-one-command-interactive-setup-detect/13-08-PLAN.md`
- **宿主 AI 配置**：Claude/Codex/Cursor/Gemini/OpenCode 等宿主的 plugin、Skill 或规则加载配置。 原因：宿主配置会改变 AI 后续工作方式，可能和用户已有规则冲突。 证据：`.claude/skills/openspec-apply-change/SKILL.md`, `.claude/skills/openspec-archive-change/SKILL.md`, `.claude/skills/openspec-explore/SKILL.md`, `.claude/skills/openspec-propose/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：`.opencode/skills/harness-init/SKILL.md`, `.opencode/skills/nano-brain/references/http-api.md`, `.planning/phases/08-session-harvest-unification-ticket-linking/08-03-PLAN.md`, `.planning/phases/13-interactive-init-wizard-one-command-interactive-setup-detect/13-08-PLAN.md`
- **宿主 AI 上下文**：AI Context Pack、Prompt Preview、Skill 路由、风险规则和项目事实。 原因：导入上下文会影响宿主 AI 后续判断，必须避免把未验证项包装成事实。

### 最小安全下一步

- **先跑 Prompt Preview**：用安装前交互式试用判断工作方式是否匹配，不需要授权或改环境。（适用：任何项目都适用，尤其是输出质量未知时。）
- **只在隔离目录或测试账号试装**：避免安装命令污染主力宿主 AI、真实项目或用户主目录。（适用：存在命令执行、插件配置或本地写入线索时。）
- **先备份宿主 AI 配置**：Skill、plugin、规则文件可能改变 Claude/Cursor/Codex 的默认行为。（适用：存在插件 manifest、Skill 或宿主规则入口时。）
- **安装后只验证一个最小任务**：先验证加载、兼容、输出质量和回滚，再决定是否深用。（适用：准备从试用进入真实工作流时。）

### 退出方式

- **保留安装前状态**：记录原始宿主配置和项目状态，后续才能判断是否可恢复。
- **准备移除宿主 plugin / Skill / 规则入口**：如果试装后行为异常，可以把宿主 AI 恢复到试装前状态。
- **记录安装命令和写入路径**：没有明确卸载说明时，至少要知道哪些目录或配置需要手动清理。
- **如果没有回滚路径，不进入主力环境**：不可回滚是继续前阻断项，不应靠信任或运气继续。

## 哪些只能预览

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

## 哪些必须安装后验证

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

## 边界与风险判断卡

- **把安装前预览误认为真实运行**：用户可能高估项目已经完成的配置、权限和兼容性验证。 处理方式：明确区分 prompt_preview_can_do 与 runtime_required。 Claim：`clm_0028` 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：`.opencode/skills/harness-init/SKILL.md`, `.opencode/skills/nano-brain/references/http-api.md`, `.planning/phases/08-session-harvest-unification-ticket-linking/08-03-PLAN.md`, `.planning/phases/13-interactive-init-wizard-one-command-interactive-setup-detect/13-08-PLAN.md` Claim：`clm_0029` supported 0.86
- **待确认**：真实安装后是否与用户当前宿主 AI 版本兼容？。原因：兼容性只能通过实际宿主环境验证。
- **待确认**：项目输出质量是否满足用户具体任务？。原因：安装前预览只能展示流程和边界，不能替代真实评测。
- **待确认**：安装命令是否需要网络、权限或全局写入？。原因：这影响企业环境和个人环境的安装风险。

## 开工前工作上下文

### 加载顺序

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

### 任务路由

- **AI Skill / Agent 指令资产库**：先基于 role_skill_index / evidence_index 帮用户挑选可用角色、Skill 或工作流。 边界：可做安装前 Prompt 体验。 证据：`.claude/skills/openspec-apply-change/SKILL.md`, `.claude/skills/openspec-archive-change/SKILL.md`, `.claude/skills/openspec-explore/SKILL.md`, `.claude/skills/openspec-propose/SKILL.md` 等 Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86
- **命令行启动或安装流程**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`.opencode/skills/harness-init/SKILL.md`, `.opencode/skills/nano-brain/references/http-api.md`, `.planning/phases/08-session-harvest-unification-ticket-linking/08-03-PLAN.md`, `.planning/phases/13-interactive-init-wizard-one-command-interactive-setup-detect/13-08-PLAN.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86

### 上下文规模

- 文件总数：1587
- 重要文件覆盖：40/1587
- 证据索引条目：80
- 角色 / Skill 条目：14

### 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 nano-brain 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 nano-brain 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 nano-brain 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 14 个角色 / Skill / 项目文档条目。

- **openspec-apply-change**（skill）：Implement tasks from an OpenSpec change. Use when the user wants to start implementing, continue implementation, or work through tasks. 激活提示：当用户任务与“openspec-apply-change”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`.claude/skills/openspec-apply-change/SKILL.md`
- **openspec-archive-change**（skill）：Archive a completed change in the experimental workflow. Use when the user wants to finalize and archive a change after implementation is complete. 激活提示：当用户任务与“openspec-archive-change”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`.claude/skills/openspec-archive-change/SKILL.md`
- **openspec-explore**（skill）：Enter explore mode - a thinking partner for exploring ideas, investigating problems, and clarifying requirements. Use when the user wants to think through something before or during a change. 激活提示：当用户任务与“openspec-explore”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`.claude/skills/openspec-explore/SKILL.md`
- **openspec-propose**（skill）：Propose a new change with all artifacts generated in one step. Use when the user wants to quickly describe what they want to build and get a complete proposal with design, specs, and tasks ready for implementation. 激活提示：当用户任务与“openspec-propose”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`.claude/skills/openspec-propose/SKILL.md`
- **Debugging Skill**（skill）：Use mode=debugging on memory search or memory query to search code, sessions, and config in a single call. 激活提示：当用户任务与“Debugging Skill”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：`.opencode/skills/debugging/SKILL.md`
- **deep-design**（skill）：MANDATORY for any non-trivial feature planning. Multi-agent pipeline: Metis scope/risk + Oracle architecture in parallel → cross-critique → confidence-scored synthesis → early Momus sanity check → OpenSpec proposal → full Momus review. MUST USE when: user describes a feature that touches multiple services or modules, user wants to plan/design/architect before coding, user says 'plan this', 'design this', 'think this… 激活提示：当用户任务与“deep-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：`.opencode/skills/deep-design/SKILL.md`
- **harness-init**（skill）： 激活提示：当用户任务与“harness-init”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：`.opencode/skills/harness-init/SKILL.md`
- **nano-brain**（skill）：Persistent memory + code intelligence for AI coding agents. Hybrid search BM25 + vector + RRF , cross-session recall, symbol graph analysis, impact checks, OpenCode/Claude Code session harvesting. Use when you need to recall prior decisions, search across sessions/codebase, trace symbol callers/callees, or persist long-term context. 激活提示：当用户任务与“nano-brain”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：`.opencode/skills/nano-brain/SKILL.md`
- **openspec-apply-change**（skill）：Implement tasks from an OpenSpec change. Use when the user wants to start implementing, continue implementation, or work through tasks. 激活提示：当用户任务与“openspec-apply-change”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：`.opencode/skills/openspec-apply-change/SKILL.md`
- **openspec-archive-change**（skill）：Archive a completed change in the experimental workflow. Use when the user wants to finalize and archive a change after implementation is complete. 激活提示：当用户任务与“openspec-archive-change”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：`.opencode/skills/openspec-archive-change/SKILL.md`
- **openspec-explore**（skill）：Enter explore mode - a thinking partner for exploring ideas, investigating problems, and clarifying requirements. Use when the user wants to think through something before or during a change. 激活提示：当用户任务与“openspec-explore”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：`.opencode/skills/openspec-explore/SKILL.md`
- **openspec-propose**（skill）：Propose a new change with all artifacts generated in one step. Use when the user wants to quickly describe what they want to build and get a complete proposal with design, specs, and tasks ready for implementation. 激活提示：当用户任务与“openspec-propose”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：`.opencode/skills/openspec-propose/SKILL.md`
- **SKILL**（skill）： 激活提示：当用户任务与“SKILL”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：`docs/SKILL.md`
- **nano-brain**（skill）：Persistent memory + code intelligence for AI coding agents. Hybrid search BM25 + vector + RRF + recency , cross-session recall, symbol analysis, impact checks, knowledge graph, OpenCode/Claude Code session harvesting. Use this skill when you need to recall prior decisions across sessions, search across multiple codebases, trace symbol callers/callees, analyze code impact what breaks if X changes , persist long-term… 激活提示：当用户任务与“nano-brain”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/nano-brain/SKILL.md`

## 证据索引

- 共索引 80 条证据。

- **Architecture Decision Records ADRs**（documentation）：Durable records of architectural and design decisions. Each ADR captures why a decision was made, what alternatives were considered, and the tradeoffs accepted. 证据：`docs/decisions/README.md`
- **Evidence**（documentation）：Screenshots, recordings, and test artifacts for high-risk changes. 证据：`docs/evidence/README.md`
- **Evidence: fix-summary-workspace-registration-leaks 238**（documentation）：Evidence: fix-summary-workspace-registration-leaks 238 证据：`docs/evidence/fix-summary-workspace-registration-leaks/README.md`
- **Vue SFC Code Intelligence Support — Design Pipeline**（documentation）：Vue SFC Code Intelligence Support — Design Pipeline 证据：`docs/evidence/vue-sfc-design/README.md`
- **Harness Gate Instructions**（documentation）：This directory contains per-gate protocol documents for the harness loop plugin. 证据：`docs/harness/gates/README.md`
- **Agent Skills**（documentation）：This directory contains skills https://docs.claude.com/en/docs/claude-code/skills that teach AI coding agents Claude Code, OpenCode, MCP-aware tools how to use nano-brain effectively. 证据：`skills/README.md`
- **Rails Extraction Benchmarks**（documentation）：End-to-end benchmarks measuring nano-brain's Ruby/Rails extraction quality and performance: route parsing, control-flow graphs CFGs , and flow builder end-to-end. 证据：`benchmarks/rails/README.md`
- **Rails Capability Benchmark**（documentation）：Static end-to-end benchmark that measures how well the live nano-brain server helps understand a Rails photo-print pipeline codebase. Each task is a hand-curated developer/support question with ground-truth symbols/files that a correct answer must surface. 证据：`benchmarks/rails/capability/README.md`
- **TypeScript Capability Benchmark**（documentation）：Agent-oriented benchmark profile for a TypeScript/JavaScript CS2 item-trading workspace. The committed dataset uses privacy-safe placeholders; local runs provide the real workspace hash through NANO BRAIN WORKSPACE . 证据：`benchmarks/typescript/capability/README.md`
- **Vue Capability Benchmark**（documentation）：Static end-to-end benchmark that measures how well the live nano-brain server helps understand a Vue.js codebase. Each task is a hand-curated developer question with ground-truth symbols/files that a correct answer must surface. 证据：`benchmarks/vue/capability/README.md`
- **SKILL**（skill_instruction）：SKILL 证据：`docs/SKILL.md`
- **CLAUDE.md**（documentation）：This file provides guidance to Claude Code claude.ai/code when working with code in this repository. 证据：`CLAUDE.md`
- **nano-brain Setup Guide for AI Agents**（documentation）：nano-brain Setup Guide for AI Agents 证据：`docs/SETUP_AGENT.md`
- **cmd/nano-brain — CLI Dispatcher + Server Entry Point**（documentation）：cmd/nano-brain — CLI Dispatcher + Server Entry Point 证据：`cmd/nano-brain/AGENTS.md`
- **internal/embed — Embedding Queue**（documentation）：Purpose: Embedding queue — async worker that generates vector embeddings via Ollama or Voyage AI. 证据：`internal/embed/AGENTS.md`
- **internal/graph — AGENTS.md**（documentation）：Code-intelligence graph layer: framework route extraction, outbound integration edges, and per-function control-flow graphs CFGs . Built on the vendored gotreesitter library Go bindings over tree-sitter grammars . 证据：`internal/graph/AGENTS.md`
- **harvest package**（documentation）：Session harvesting — ingests OpenCode and Claude Code AI sessions into nano-brain storage. 证据：`internal/harvest/AGENTS.md`
- **internal/mcp — MCP Protocol Server**（documentation）：MCP Model Context Protocol server — 16 tools for AI agent memory and code intelligence access. 证据：`internal/mcp/AGENTS.md`
- **internal/search**（documentation）：Hybrid search pipeline: BM25 + vector + RRF fusion + recency decay + reranking. 证据：`internal/search/AGENTS.md`
- **handlers**（documentation）：HTTP handler functions for all REST API endpoints. One file per endpoint group. 证据：`internal/server/handlers/AGENTS.md`
- **storage — AGENTS.md**（documentation）：PostgreSQL storage layer — pool management, sqlc-generated queries, goose migrations. 证据：`internal/storage/AGENTS.md`
- **Package**（package_manifest）：{ "name": "@nano-step/nano-brain", "version": "0.0.0-dev", "description": "Persistent memory and code intelligence for AI coding agents", "bin": { "nano-brain": "npm/run.js" }, "scripts": { "postinstall": "node npm/postinstall.js" }, "publishConfig": { "access": "public" }, "repository": { "type": "git", "url": "git+https://github.com/nano-step/nano-brain.git" }, "keywords": "ai", "memory", "mcp", "coding-agent", "search", "pgvector" , "author": "nano-step", "license": "MIT", "files": "npm/", "README.md", "LICENSE" , "dependencies": { "@nano-step/oh-my-harness": "^1.1.3" } } 证据：`package.json`
- **nano-brain**（skill_instruction）：Persistent memory for AI coding agents. Hybrid search BM25 + semantic + LLM reranking across past sessions, codebase, notes, and daily logs. 证据：`SKILL.md`
- **Implementing: schema:**（skill_instruction）：Implement tasks from an OpenSpec change. 证据：`.claude/skills/openspec-apply-change/SKILL.md`
- **Archive Complete**（skill_instruction）：Archive a completed change in the experimental workflow. 证据：`.claude/skills/openspec-archive-change/SKILL.md`
- **The Stance**（skill_instruction）：Enter explore mode. Think deeply. Visualize freely. Follow the conversation wherever it goes. 证据：`.claude/skills/openspec-explore/SKILL.md`
- **Skill**（skill_instruction）：Propose a new change - create the change and generate all artifacts in one step. 证据：`.claude/skills/openspec-propose/SKILL.md`
- **Debugging Skill**（skill_instruction）：Use mode=debugging on memory search or memory query to search code, sessions, and config in a single call. 证据：`.opencode/skills/debugging/SKILL.md`
- **Deep Design Pipeline**（skill_instruction）：A 7-phase orchestration pipeline that turns a user's feature idea into a reviewed, implementation-ready OpenSpec proposal. Inspired by the LLM Council pattern: agents analyze independently, then cross-critique each other's outputs, conflicts are resolved using confidence scoring not hard rules , and a sanity-check gate catches design flaws before expensive spec-writing begins. 证据：`.opencode/skills/deep-design/SKILL.md`
- **Harness Init**（skill_instruction）：Generate a production-ready engineering harness in the current project. 证据：`.opencode/skills/harness-init/SKILL.md`
- **nano-brain**（skill_instruction）：Persistent memory + code-intel daemon. Agents talk to it via MCP — streamable HTTP at /mcp on the daemon. HTTP is available as a fallback for scripts and integration tests. 证据：`.opencode/skills/nano-brain/SKILL.md`
- **Implementing: schema:**（skill_instruction）：Implement tasks from an OpenSpec change. 证据：`.opencode/skills/openspec-apply-change/SKILL.md`
- **Archive Complete**（skill_instruction）：Archive a completed change in the experimental workflow. 证据：`.opencode/skills/openspec-archive-change/SKILL.md`
- **The Stance**（skill_instruction）：Enter explore mode. Think deeply. Visualize freely. Follow the conversation wherever it goes. 证据：`.opencode/skills/openspec-explore/SKILL.md`
- **Skill**（skill_instruction）：Propose a new change - create the change and generate all artifacts in one step. 证据：`.opencode/skills/openspec-propose/SKILL.md`
- **nano-brain — Persistent Memory + Code Intelligence**（skill_instruction）：nano-brain — Persistent Memory + Code Intelligence 证据：`skills/nano-brain/SKILL.md`
- **Package**（package_manifest）：{"dependencies":{"nuxt":"^3.0.0"}} 证据：`internal/graph/testdata/nuxt-fixture/package.json`
- **.opencode/harness.config.json Reference**（documentation）：.opencode/harness.config.json Reference 证据：`.opencode/harness.config.README.md`
- **AGENTS SNIPPET**（documentation）：AGENTS SNIPPET 证据：`docs/AGENTS_SNIPPET.md`
- **Context Retrieval Rules**（documentation）：Agents working in this harness MUST read only the files listed below for each phase × lane combination. Reading more is token waste . Reading less risks missing required rules. 证据：`docs/CONTEXT_RULES.md`
- **Action plan — dashboard split executable backlog**（documentation）：Action plan — dashboard split executable backlog 证据：`docs/DASHBOARD_SPLIT_ACTION_PLAN.md`
- **Plan — Split the dashboard into its own repo; nano-brain becomes API-only**（documentation）：Plan — Split the dashboard into its own repo; nano-brain becomes API-only 证据：`docs/DASHBOARD_SPLIT_PLAN.md`
- **Feature Showcase**（documentation）：Visual examples of nano-brain's capabilities. 证据：`docs/FEATURES.md`
- **Feature Intake**（documentation）：Every implementation prompt enters the intake gate before code changes. The human does not need to classify risk. The harness does. 证据：`docs/FEATURE_INTAKE.md`
- **Migration Plan — Unify graph visualization on AntV G6**（documentation）：Migration Plan — Unify graph visualization on AntV G6 证据：`docs/G6_MIGRATION_PLAN.md`
- **Glossary**（documentation）：Closed-set vocabulary for the nano-brain harness. When a term appears in any harness doc HARNESS.md , HARNESS GATES.md , FEATURE INTAKE.md , ADRs, story templates it carries the meaning defined here — no other meaning is permitted. 证据：`docs/GLOSSARY.md`
- **Harness Backlog**（documentation）：Use this file when an agent discovers a missing harness capability but should not change the operating model immediately. 证据：`docs/HARNESS_BACKLOG.md`
- **Harness Gates — nano-brain v2**（documentation）：Gate specification for the nano-brain v2 Go project. Each gate defines the checks required before proceeding to the next phase of development. 证据：`docs/HARNESS_GATES.md`
- **Harness Runner Contract**（documentation）：Status: legacy interface. The autonomous gate-loop plugin that consumed this contract was OpenCode-only .opencode/plugin/harness-loop/ , /harness-on and is no longer the driver. On Claude Code, gates are driven by the /harness-gsd command autonomous pipeline or the harness-check skill manual runs — see docs/HARNESS.md § Entry points. This document remains the authoritative spec for scripts/harness-check.sh --json output, which any driver hook, CI, future plugin can consume. "Plugin" below refers to the legacy OpenCode consumer. 证据：`docs/HARNESS_RUNNER_CONTRACT.md`
- **nano-brain Roadmap**（documentation）：nano-brain is a persistent memory and code intelligence layer for AI coding agents. Goal: agents know the project context, decision history, and can anticipate what's needed next — across sessions, machines, and team members. 证据：`docs/ROADMAP.md`
- **Trace Specification**（documentation）：A trace is a structured record of what an agent did during a task. Traces feed Retro Gate ⑥ metric collection and serve as institutional memory for future agents. 证据：`docs/TRACE_SPEC.md`
- **nano-brain v2 — System Architecture Diagram**（documentation）：nano-brain v2 — System Architecture Diagram 证据：`docs/architecture-diagram.md`
- **Architecture Decision Document**（documentation）：This document builds collaboratively through step-by-step discovery. Sections are appended as we work through each architectural decision together. 证据：`docs/architecture.md`
- **nano-brain-greenfield - Epic Breakdown**（documentation）：nano-brain-greenfield - Epic Breakdown 证据：`docs/epics.md`
- **Harness Philosophy**（documentation）：The app is what users touch. The harness is what agents touch. 证据：`.opencode/skills/harness-init/docs/PHILOSOPHY.md`
- **Decision Log — nano-brain v2 brief**（documentation）：2026-05-22T05:30:00Z — Run started - Branch: b-main commit e561fac - Facilitator: Sisyphus orchestrator - User: BMad - Communication language: Vietnamese - Document output language: English - Reference inputs available: docs/reference-prd.md, docs/reference-readme.md - Source materials beyond references: none 证据：`docs/briefs/brief-nano-brain-greenfield-2026-05-22/.decision-log.md`
- **Product Brief: nano-brain v2**（documentation）：Status: Discovery complete. Brief ready for review. North Star locked 2026-05-22 : Reliability / correctness. Other dimensions are secondary. Recreation thesis locked 2026-05-22 : Full rewrite of nano-brain — same feature set, new architecture, no refactor. 证据：`docs/briefs/brief-nano-brain-greenfield-2026-05-22/brief.md`
- **Issue 234 Fix Verification**（documentation）：Issue : CLI HTTP client ignores --config / NANO BRAIN CONFIG for server host/port — only reads NANO BRAIN HOST / NANO BRAIN PORT env vars 证据：`docs/evidence/234/fix-verification.md`
- **Issue 234 Fix Verification**（documentation）：Issue : CLI HTTP client ignores --config / NANO BRAIN CONFIG for server host/port — only reads NANO BRAIN HOST / NANO BRAIN PORT env vars 证据：`docs/evidence/234/self-review-234.md`
- **Issue 235 Fix Verification**（documentation）：Issue : CLI --json output mixes zerolog INFO lines with response on stdout, breaking jq pipelines 证据：`docs/evidence/235/fix-verification.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/decisions/README.md`, `docs/evidence/README.md`, `docs/evidence/fix-summary-workspace-registration-leaks/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/decisions/README.md`, `docs/evidence/README.md`, `docs/evidence/fix-summary-workspace-registration-leaks/README.md`

## 用户开工前应该回答的问题

- 你准备在哪个宿主 AI 或本地环境中使用它？
- 你只是想先体验工作流，还是准备真实安装？
- 你最在意的是安装成本、输出质量、还是和现有规则的冲突？

## 验收标准

- 所有能力声明都能回指到 evidence_refs 中的文件路径。
- AI_CONTEXT_PACK.md 没有把预览包装成真实运行。
- 用户能在 3 分钟内看懂适合谁、能做什么、如何开始和风险边界。

---

## Doramagic Context Augmentation

下面内容用于强化 Repomix/AI Context Pack 主体。Human Manual 只提供阅读骨架；踩坑日志会被转成宿主 AI 必须遵守的工作约束。

## Human Manual 骨架

使用规则：这里只是项目阅读路线和显著性信号，不是事实权威。具体事实仍必须回到 repo evidence / Claim Graph。

宿主 AI 硬性规则：
- 不得把页标题、章节顺序、摘要或 importance 当作项目事实证据。
- 解释 Human Manual 骨架时，必须明确说它只是阅读路线/显著性信号。
- 能力、安装、兼容性、运行状态和风险判断必须引用 repo evidence、source path 或 Claim Graph。

- **项目概览与系统架构**：importance `high`
  - source_paths: README.md, cmd/nano-brain/main.go, internal/server/routes.go, internal/server/server.go, internal/mcp/server.go
- **混合检索与代码智能（MCP 工具集）**：importance `high`
  - source_paths: internal/mcp/tools.go, internal/mcp/streamable.go, internal/mcp/graph_paths.go, internal/search/search.go, internal/search/rrf.go
- **数据管理：会话采集、索引与存储流水线**：importance `high`
  - source_paths: internal/harvest/engine.go, internal/harvest/runner.go, internal/harvest/opencode_sqlite.go, internal/harvest/claudecode.go, internal/codesummarize/service.go
- **部署、配置、CLI 与 REST API**：importance `high`
  - source_paths: install.sh, cmd/nano-brain/init.go, cmd/nano-brain/init_serve.go, cmd/nano-brain/mcp_client_config.go, cmd/nano-brain/doctor.go

## Repo Inspection Evidence / 源码检查证据

- repo_clone_verified: true
- repo_inspection_verified: true
- repo_commit: `9764ba6bb17b35b5c53026342291bc4dde5e40b0`
- inspected_files: `Dockerfile`, `README.md`, `docker-compose.yml`, `package.json`, `docs/AGENTS_SNIPPET.md`, `docs/CONFIGURATION.md`, `docs/CONTEXT_RULES.md`, `docs/DASHBOARD_SPLIT_ACTION_PLAN.md`, `docs/DASHBOARD_SPLIT_JUNIOR_GUIDE.md`, `docs/DASHBOARD_SPLIT_PLAN.md`, `docs/FEATURES.md`, `docs/FEATURE_INTAKE.md`, `docs/G6_MIGRATION_PLAN.md`, `docs/GLOSSARY.md`, `docs/HARNESS.md`, `docs/HARNESS_BACKLOG.md`, `docs/HARNESS_GATES.md`, `docs/HARNESS_RUNNER_CONTRACT.md`, `docs/ROADMAP.md`, `docs/SETUP_AGENT.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: 失败模式：installation: Interactive MCP client auto-configuration after workspace registration

- Trigger: Developers should check this installation risk before relying on the project: Interactive MCP client auto-configuration after workspace registration
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: Interactive MCP client auto-configuration after workspace registration. Context: Observed during installation or first-run setup.
- Why it matters: Developers may fail before the first successful local run: Interactive MCP client auto-configuration after workspace registration
- Evidence: failure_mode_cluster:github_issue | https://github.com/nano-step/nano-brain/issues/525 | Interactive MCP client auto-configuration after workspace registration
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 2: 失败模式：installation: v2026.7.0303

- Trigger: Developers should check this installation risk before relying on the project: v2026.7.0303
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: v2026.7.0303. Context: Observed when using node
- Why it matters: Upgrade or migration may change expected behavior: v2026.7.0303
- Evidence: failure_mode_cluster:github_release | https://github.com/nano-step/nano-brain/releases/tag/v2026.7.0303 | v2026.7.0303
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 3: 来源证据：Code-graph extractor bypasses document-indexer file filters (.gitignore, max_file_size, minified) → orphan edges poison…

- Trigger: GitHub 社区证据显示该项目存在一个安装相关的待验证问题：Code-graph extractor bypasses document-indexer file filters (.gitignore, max_file_size, minified) → orphan edges poison memory_trace/graph/symbols for agents
- Host AI rule: 来源显示可能已有修复、规避或版本变化，说明书中必须标注适用版本。
- Why it matters: 可能增加新用户试用和生产接入成本。
- Evidence: community_evidence:github | https://github.com/nano-step/nano-brain/issues/535 | 来源讨论提到 node 相关条件，需在安装/试用前复核。
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 4: 来源证据：Interactive MCP client auto-configuration after workspace registration

- Trigger: GitHub 社区证据显示该项目存在一个安装相关的待验证问题：Interactive MCP client auto-configuration after workspace registration
- Host AI rule: 来源显示可能已有修复、规避或版本变化，说明书中必须标注适用版本。
- Why it matters: 可能增加新用户试用和生产接入成本。
- Evidence: community_evidence:github | https://github.com/nano-step/nano-brain/issues/525 | 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 5: 来源证据：MCP: bind default workspace via URL query param, skip agent-side discovery

- Trigger: GitHub 社区证据显示该项目存在一个安装相关的待验证问题：MCP: bind default workspace via URL query param, skip agent-side discovery
- Host AI rule: 来源显示可能已有修复、规避或版本变化，说明书中必须标注适用版本。
- Why it matters: 可能增加新用户试用和生产接入成本。
- Evidence: community_evidence:github | https://github.com/nano-step/nano-brain/issues/523 | 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 6: 来源证据：watcher: gate code-summarize notify on files that actually yield symbols

- Trigger: GitHub 社区证据显示该项目存在一个安装相关的待验证问题：watcher: gate code-summarize notify on files that actually yield symbols
- Host AI rule: 来源显示可能已有修复、规避或版本变化，说明书中必须标注适用版本。
- Why it matters: 可能增加新用户试用和生产接入成本。
- Evidence: community_evidence:github | https://github.com/nano-step/nano-brain/issues/520 | 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 7: 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/nano-step/nano-brain | 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 8: 失败模式：configuration: Add OpenAPI 3.0 spec for the REST API, so non-MCP clients can discover all endpoints

- Trigger: Developers should check this configuration risk before relying on the project: Add OpenAPI 3.0 spec for the REST API, so non-MCP clients can discover all endpoints
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: Add OpenAPI 3.0 spec for the REST API, so non-MCP clients can discover all endpoints. Context: Observed during installation or first-run setup.
- Why it matters: Developers may misconfigure credentials, environment, or host setup: Add OpenAPI 3.0 spec for the REST API, so non-MCP clients can discover all endpoints
- Evidence: failure_mode_cluster:github_issue | https://github.com/nano-step/nano-brain/issues/530 | Add OpenAPI 3.0 spec for the REST API, so non-MCP clients can discover all endpoints
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 9: 失败模式：configuration: CLI --help is misleading: -h/--help shows nothing useful, no-args starts a real server, help...

- Trigger: Developers should check this configuration risk before relying on the project: CLI --help is misleading: -h/--help shows nothing useful, no-args starts a real server, help list is stale
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: CLI --help is misleading: -h/--help shows nothing useful, no-args starts a real server, help list is stale. Context: Source discussion did not expose a precise runtime context.
- Why it matters: Developers may misconfigure credentials, environment, or host setup: CLI --help is misleading: -h/--help shows nothing useful, no-args starts a real server, help list is stale
- Evidence: failure_mode_cluster:github_issue | https://github.com/nano-step/nano-brain/issues/527 | CLI --help is misleading: -h/--help shows nothing useful, no-args starts a real server, help list is stale
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 10: 失败模式：configuration: MCP: bind default workspace via URL query param, skip agent-side discovery

- Trigger: Developers should check this configuration risk before relying on the project: MCP: bind default workspace via URL query param, skip agent-side discovery
- Host AI rule: Before packaging this project, run the relevant install/config/quickstart check for: MCP: bind default workspace via URL query param, skip agent-side discovery. Context: Observed during installation or first-run setup.
- Why it matters: Developers may misconfigure credentials, environment, or host setup: MCP: bind default workspace via URL query param, skip agent-side discovery
- Evidence: failure_mode_cluster:github_issue | https://github.com/nano-step/nano-brain/issues/523 | MCP: bind default workspace via URL query param, skip agent-side discovery, failure_mode_cluster:github_issue | https://github.com/nano-step/nano-brain/issues/522 | MCP: bind default workspace via URL query param, skip agent-side discovery
- 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。