# knitbrain - 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 knitbrain 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_0002` supported 0.86

## 它能做什么

- **命令行启动或安装流程**（需要安装后验证）：项目文档中存在可执行命令，真实使用需要在本地或宿主环境中运行这些命令。 证据：`README.md` Claim：`clm_0001` supported 0.86

## 怎么开始

- `npm install -g knitbrain      # or: npx knitbrain <command>` 证据：`README.md` Claim：`clm_0003` supported 0.86

## 继续前判断卡

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

### 30 秒判断

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

### 现在可以相信

- **适合人群线索：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_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：`README.md` Claim：`clm_0001` 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_0003` 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.md`
- **可安全回滚不能默认相信。**（unverified）：除非项目明确提供卸载和恢复说明，否则必须先在隔离环境验证。
- **真实安装后是否与用户当前宿主 AI 版本兼容？**（unverified）：兼容性只能通过实际宿主环境验证。
- **项目输出质量是否满足用户具体任务？**（unverified）：安装前预览只能展示流程和边界，不能替代真实评测。
- **安装命令是否需要网络、权限或全局写入？**（unverified）：这影响企业环境和个人环境的安装风险。 证据：`README.md`

### 继续会触碰什么

- **命令执行**：包管理器、网络下载、本地插件目录、项目配置或用户主目录。 原因：运行第一条命令就可能产生环境改动；必须先判断是否值得跑。 证据：`README.md`
- **宿主 AI 配置**：Claude/Codex/Cursor/Gemini/OpenCode 等宿主的 plugin、Skill 或规则加载配置。 原因：宿主配置会改变 AI 后续工作方式，可能和用户已有规则冲突。 证据：`CLAUDE.md`
- **本地环境或项目文件**：安装结果、插件缓存、项目配置或本地依赖目录。 原因：安装前无法证明写入范围和回滚方式，需要隔离验证。 证据：`README.md`
- **宿主 AI 上下文**：AI Context Pack、Prompt Preview、Skill 路由、风险规则和项目事实。 原因：导入上下文会影响宿主 AI 后续判断，必须避免把未验证项包装成事实。

### 最小安全下一步

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

### 退出方式

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

## 哪些只能预览

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

## 哪些必须安装后验证

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

## 边界与风险判断卡

- **把安装前预览误认为真实运行**：用户可能高估项目已经完成的配置、权限和兼容性验证。 处理方式：明确区分 prompt_preview_can_do 与 runtime_required。 Claim：`clm_0004` 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_0005` 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

- **命令行启动或安装流程**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`README.md` Claim：`clm_0001` supported 0.86

### 上下文规模

- 文件总数：80
- 重要文件覆盖：40/80
- 证据索引条目：58
- 角色 / Skill 条目：12

### 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 knitbrain 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 knitbrain 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 knitbrain 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 12 个角色 / Skill / 项目文档条目。

- **knitbrain**（project_doc）：typescript project. Knit-powered workflow. The protocol depth is fetched on demand via knit get workflow {phase} — this file holds only project-specific facts. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`CLAUDE.md`
- **Install**（project_doc）：The local-first brain for coding agents. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`README.md`
- **knitbrain roadmap — b live agent CRM · c parallel orchestration**（project_doc）：knitbrain roadmap — b live agent CRM · c parallel orchestration 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/ROADMAP.md`
- **architect-reviewer**（project_doc）：Use this agent when you need to evaluate system design decisions, architectural patterns, and technology choices at the macro level. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`.claude/agents/knit-architect-reviewer.md`
- **build-engineer**（project_doc）：Use this agent when you need to optimize build performance, reduce compilation times, or scale build systems across growing teams. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`.claude/agents/knit-build-engineer.md`
- **code-reviewer**（project_doc）：Use this agent when you need to conduct comprehensive code reviews focusing on code quality, security vulnerabilities, and best practices. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`.claude/agents/knit-code-reviewer.md`
- **debugger**（project_doc）：Use this agent when you need to diagnose and fix bugs, identify root causes of failures, or analyze error logs and stack traces to resolve issues. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`.claude/agents/knit-debugger.md`
- **qa-expert**（project_doc）：Use this agent when you need comprehensive quality assurance strategy, test planning across the entire development cycle, or quality metrics analysis to improve overall software quality. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`.claude/agents/knit-qa-expert.md`
- **typescript-pro**（project_doc）：Use when implementing TypeScript code requiring advanced type system patterns, complex generics, type-level programming, or end-to-end type safety across full-stack applications. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`.claude/agents/knit-typescript-pro.md`
- **Meter**（project_doc）：Show the Knit Brain context-window meter usage %, tokens saved, handoff advice 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`.claude/commands/meter.md`
- **Knit Brain**（project_doc）：Knit Brain compresses large tool outputs into skeletons. A ⟨ccr:HASH⟩ marker means the exact original is stored locally — call the knitbrain retrieve tool with that hash to read it byte-for-byte. Check knitbrain context meter periodically; when it says to, save a handoff with knitbrain save handoff and start a fresh session knitbrain load session restores everything . When the user states a task, call knitbrain run… 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`.claude/rules/knitbrain.md`
- **Terse mode output tokens**（project_doc）：Knit Brain compresses large tool outputs into skeletons. A ⟨ccr:HASH⟩ marker means the exact original is stored locally — call the knitbrain retrieve tool with that hash to read it byte-for-byte. Check knitbrain context meter periodically; when it says to, save a handoff with knitbrain save handoff and start a fresh session knitbrain load session restores everything . When the user states a task, call knitbrain run… 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`.windsurf/rules/knitbrain.md`

## 证据索引

- 共索引 58 条证据。

- **knitbrain**（documentation）：typescript project. Knit-powered workflow. The protocol depth is fetched on demand via knit get workflow {phase} — this file holds only project-specific facts. 证据：`CLAUDE.md`
- **Install**（documentation）：The local-first brain for coding agents. 证据：`README.md`
- **Package**（package_manifest）：{ "name": "knitbrain", "version": "0.4.4", "description": "The local-first brain for coding agents: per-project memory, task-tier workflow routing, and lossless context compression — measured ~48% of all tool-result tokens on real sessions 60–70% on code/JSON/logs , answer-preservation gated, reproducible with one command.", "type": "module", "main": "dist/lib.js", "types": "dist/lib.d.ts", "exports": { ".": { "types": "./dist/lib.d.ts", "default": "./dist/lib.js" } }, "bin": { "knitbrain": "dist/index.js", "knitbrain-proxy": "dist/proxy/index.js", "knitbrain-hook": "dist/hooks/index.js" }, "keywords": "mcp", "model-context-protocol", "token-optimization", "context-compression", "llm", "ai-… 证据：`package.json`
- **License**（source_file）：Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files the "Software" , to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: 证据：`LICENSE`
- **knitbrain roadmap — b live agent CRM · c parallel orchestration**（documentation）：knitbrain roadmap — b live agent CRM · c parallel orchestration 证据：`docs/ROADMAP.md`
- **Tsconfig.Build**（structured_config）：{ "extends": "./tsconfig.json", "compilerOptions": { "rootDir": "src" }, "include": "src/ / .ts" , "exclude": "dist", "node modules", "tests", "scripts" } 证据：`tsconfig.build.json`
- **Bench**（source_file）：import { mkdtempSync, rmSync } from "node:fs"; import { tmpdir } from "node:os"; import { join } from "node:path"; import { activeTokenizerName } from "../src/tokenizer.js"; import { measure, summarize, type Measurement } from "../src/measure.js"; import { createFileCCRStore } from "../src/ccr/store.js"; import { compressJson } from "../src/optimizer/json.js"; import { compressCode } from "../src/optimizer/code.js"; import { compress } from "../src/optimizer/router.js"; import { ensureAst } from "../src/optimizer/ast.js"; import { IMPORTANT LINE } from "../src/optimizer/structured.js"; function codeShape : string ⋮---- const fn = i: number : string = / Handles the ${i}-th request phase, val… 证据：`scripts/bench.ts`
- **Setup**（source_file）：import { existsSync } from "node:fs"; import { homedir } from "node:os"; import { join } from "node:path"; import { applyArtifacts, claudeArtifacts, codexSnippet, cursorArtifacts, vscodeArtifacts, windsurfArtifacts, universalArtifacts, windsurfSnippet, zedSnippet, copilotSnippet, } from "./platforms.js"; import { applyGlobalConfig, type GlobalConfigKind } from "./global-config.js"; export type Platform = "claude-code" "cursor" "codex" "vscode" "windsurf" "zed" "copilot-cli" "unknown"; export interface DetectInputs { env: NodeJS.ProcessEnv; exists: path: string = boolean; home: string; } export function detectPlatforms inp: DetectInputs : Platform export interface SetupConfig { mcpServers: R… 证据：`src/setup.ts`
- **Agents**（source_file）：import { mkdirSync } from "node:fs"; import { writeAtomic } from "../atomic.js"; import { dirname, join } from "node:path"; ⋮---- export interface DomainProposal { name: string; scope: string; files: string ; tools: string ; reviewGate: boolean; contextBudget: number; } export interface AgentSpec { name: string; description?: string; scope?: string; tools?: string ; reviewGate?: boolean; contextBudget?: number; brief?: string; } ⋮---- export function proposeAgents files: string : DomainProposal export function generateAgentMarkdown spec: AgentSpec : string / Write a generated agent to the project's .claude/agents directory. / export function writeAgent projectRoot: string, spec: AgentSpec :… 证据：`src/engine/agents.ts`
- **Calibration**（source_file）：import { existsSync, mkdirSync, readFileSync } from "node:fs"; import { writeAtomic } from "../atomic.js"; import { join } from "node:path"; import type { Tier } from "./workflow.js"; export interface CalibrationState { fpDirections: Record ; scopeAdjust: number; updatedAt: string; } export interface Calibration { recordFalsePositive claimed: Tier, actual: Tier : CalibrationState & { shifted: boolean }; get : CalibrationState; reset : CalibrationState; } ⋮---- recordFalsePositive claimed: Tier, actual: Tier : CalibrationState & get : CalibrationState; reset : CalibrationState; ⋮---- const fresh = : CalibrationState = export function createCalibration root: string : Calibration ⋮---- const l… 证据：`src/engine/calibration.ts`
- **Feedback**（source_file）：import { existsSync, mkdirSync, readFileSync } from "node:fs"; import { writeAtomic } from "../atomic.js"; import { join } from "node:path"; import type { ContentType } from "../optimizer/types.js"; export interface FeedbackStat { kind: ContentType; compressions: number; retrievals: number; rate: number; skipping: boolean; } export interface Feedback { onCompress kind: ContentType, handle: string : void; onRetrieve handle: string : void; shouldSkip kind: ContentType : boolean; stats : FeedbackStat ; } ⋮---- onCompress kind: ContentType, handle: string : void; onRetrieve handle: string : void; shouldSkip kind: ContentType : boolean; stats : FeedbackStat ; ⋮---- interface State { kinds: Recor… 证据：`src/engine/feedback.ts`
- **Knowledge**（source_file）：import { existsSync, mkdirSync, readdirSync, readFileSync, statSync } from "node:fs"; import { writeAtomic } from "../atomic.js"; import { dirname, join, relative, resolve } from "node:path"; export interface ImportEdge { from: string; names: string ; } export interface FileNode { file: string; imports: ImportEdge ; exports: string ; } export interface Knowledge { scan : { files: number }; queryImports file: string : ImportEdge null; queryExports file: string : string null; queryDependents file: string : string ; listFiles : string ; } ⋮---- scan : queryImports file: string : ImportEdge null; queryExports file: string : string null; queryDependents file: string : string ; listFiles : string… 证据：`src/engine/knowledge.ts`
- **Memory**（source_file）：import { createHash } from "node:crypto"; import { existsSync, mkdirSync, readFileSync, rmSync } from "node:fs"; import { writeAtomic } from "../atomic.js"; ⋮---- import { join } from "node:path"; export interface Learning { id: string; date: string; summary: string; lesson: string; tags: string ; helpful: number; unhelpful: number; } export interface LearningHeadline { id: string; summary: string; tags: string ; score: number; net: number; } export function learningHealth l: Learning : "unproven" "proven" "discredited" export interface Memory { recordLearning input: { summary: string; lesson: string; tags?: string } : { id: string; duplicate: boolean; }; searchLearnings query: string, limi… 证据：`src/engine/memory.ts`
- **Meter**（source_file）：import { existsSync, mkdirSync, readFileSync } from "node:fs"; import { writeAtomic } from "../atomic.js"; import { join } from "node:path"; export interface MeterReading { usedTokens: number; windowTokens: number; usedPct: number; savedTokens: number; status: "ok" "warn" "handoff"; advice: string; } export interface Meter { onRequest originalTokens: number, optimizedTokens: number : void; onToolOutput tokens: number : void; onSaved tokens: number : void; read : MeterReading; reset : void; } ⋮---- onRequest originalTokens: number, optimizedTokens: number : void; onToolOutput tokens: number : void; onSaved tokens: number : void; read : MeterReading; reset : void; ⋮---- export interface Meter… 证据：`src/engine/meter.ts`
- **Quota**（source_file）：import { existsSync, readFileSync } from "node:fs"; import { homedir } from "node:os"; import { join } from "node:path"; export interface QuotaWindow { label: string; usedPct: number; used: number; limit: number; resetsInMin?: number; } export interface PlatformQuota { platform: string; windows: QuotaWindow ; fetchedAt: string; } ⋮---- export function readClaudeToken home: string = homedir : string null interface RawWindow { used?: number; limit?: number; utilization?: number; resets at?: string number; } ⋮---- function resetsInMin resets at: string number undefined : number undefined export function parseClaudeUsage data: unknown : QuotaWindow export async function fetchClaudeQuota home: s… 证据：`src/engine/quota.ts`
- **Workflow**（source_file）：export type Tier = "inquiry" "trivial" "standard" "complex"; export interface Classification { tier: Tier; phases: string ; autoPlanMode: boolean; reason: string; } ⋮---- export function classifyTask description: string, files: string = , scopeAdjust = 0 : Classification 证据：`src/engine/workflow.ts`
- **Tools**（source_file）：import type { CCRStore } from "../ccr/store.js"; import { learningHealth, type Memory } from "../engine/memory.js"; import type { Knowledge } from "../engine/knowledge.js"; import type { Feedback } from "../engine/feedback.js"; import type { TeamBoard } from "../engine/teams.js"; import type { Meter } from "../engine/meter.js"; import type { SkillsStore } from "../engine/skills.js"; import { classifyTask, type Tier } from "../engine/workflow.js"; import type { Calibration } from "../engine/calibration.js"; import type { ActivityLog } from "../engine/activity.js"; import { proposeAgents, writeAgent } from "../engine/agents.js"; import { skillHealth } from "../engine/skills.js"; import { load… 证据：`src/mcp/tools.ts`
- **Router**（source_file）：import type { CCRStore } from "../ccr/store.js"; import type { ContentType } from "./types.js"; import { PARAMS } from "./params.js"; import { isJson, compressJson } from "./json.js"; import { isCode, compressCode } from "./code.js"; import { compressCodeAst } from "./ast.js"; import { compressText, compressShortProse } from "./text.js"; import { isDiff, isSearchResults, isLogOutput, compressDiff, compressSearchResults, compressLog, IMPORTANT LINE, RESULT LINE, DECLARATION LINE, } from "./structured.js"; import type { CompressResult } from "./types.js"; import { countTokens } from "../tokenizer.js"; export interface RouteResult { readonly skeleton: string; readonly handle: string; readonly… 证据：`src/optimizer/router.ts`
- **Structured**（source_file）：import type { CCRStore } from "../ccr/store.js"; import type { CompressResult } from "./types.js"; ⋮---- export function isSearchResults text: string : boolean export function compressSearchResults original: string, ccr: CCRStore : CompressResult ⋮---- const flushDrops = : void = ⋮---- export function isLogOutput text: string : boolean ⋮---- export function compressLog original: string, ccr: CCRStore : CompressResult ⋮---- const flushRun = : void = ⋮---- export function isDiff text: string : boolean / Hunk bodies longer than this elide to a ±count marker. / ⋮---- / Diff skeleton: keep file headers and hunk headers the WHERE , keep error lines, elide long hunk bodies to ⟪… +a/-b/~c lines⟫ th… 证据：`src/optimizer/structured.ts`
- **Types**（source_file）：export type ContentType = "json" "code" "text" "prose" "search" "log" "diff"; export interface CompressResult { readonly skeleton: string; readonly handle: string; readonly contentType: ContentType; } 证据：`src/optimizer/types.ts`
- **Optimize Request**（source_file）：import type { CCRStore } from "../ccr/store.js"; import { sha256 } from "../ccr/store.js"; import { compress } from "../optimizer/router.js"; import type { ContentType } from "../optimizer/types.js"; import { countTokens } from "../tokenizer.js"; import { alignDynamicContent, prefixHash } from "./cache-aligner.js"; export interface ContentBlock { type: string; text?: string; k: string : unknown; } export type MessageContent = string ContentBlock ; export interface Message { role: string; content: MessageContent; k: string : unknown; } export interface RequestBody { system?: string ContentBlock ; messages: Message ; k: string : unknown; } export interface OptimizeOptions { keepLastTurns?: nu… 证据：`src/proxy/optimize-request.ts`
- **Communication Protocol**（documentation）：You are a senior architecture reviewer with expertise in evaluating system designs, architectural decisions, and technology choices. Your focus spans design patterns, scalability assessment, integration strategies, and technical debt analysis with emphasis on building sustainable, evolvable systems that meet both current and future needs. 证据：`.claude/agents/knit-architect-reviewer.md`
- **Communication Protocol**（documentation）：You are a senior build engineer with expertise in optimizing build systems, reducing compilation times, and maximizing developer productivity. Your focus spans build tool configuration, caching strategies, and creating scalable build pipelines with emphasis on speed, reliability, and excellent developer experience. 证据：`.claude/agents/knit-build-engineer.md`
- **Communication Protocol**（documentation）：You are a senior code reviewer with expertise in identifying code quality issues, security vulnerabilities, and optimization opportunities across multiple programming languages. Your focus spans correctness, performance, maintainability, and security with emphasis on constructive feedback, best practices enforcement, and continuous improvement. 证据：`.claude/agents/knit-code-reviewer.md`
- **Communication Protocol**（documentation）：You are a senior debugging specialist with expertise in diagnosing complex software issues, analyzing system behavior, and identifying root causes. Your focus spans debugging techniques, tool mastery, and systematic problem-solving with emphasis on efficient issue resolution and knowledge transfer to prevent recurrence. 证据：`.claude/agents/knit-debugger.md`
- **Communication Protocol**（documentation）：You are a senior QA expert with expertise in comprehensive quality assurance strategies, test methodologies, and quality metrics. Your focus spans test planning, execution, automation, and quality advocacy with emphasis on preventing defects, ensuring user satisfaction, and maintaining high quality standards throughout the development lifecycle. 证据：`.claude/agents/knit-qa-expert.md`
- **Communication Protocol**（documentation）：You are a senior TypeScript developer with mastery of TypeScript 5.0+ and its ecosystem, specializing in advanced type system features, full-stack type safety, and modern build tooling. Your expertise spans frontend frameworks, Node.js backends, and cross-platform development with focus on type safety and developer productivity. 证据：`.claude/agents/knit-typescript-pro.md`
- **Meter**（documentation）：Call the knitbrain context meter tool and present the reading clearly: usage %, tokens saved this session, and the advice line. If status is "handoff", follow the advice now. 证据：`.claude/commands/meter.md`
- **Knit Brain**（documentation）：Knit Brain compresses large tool outputs into skeletons. A ⟨ccr:HASH⟩ marker means the exact original is stored locally — call the knitbrain retrieve tool with that hash to read it byte-for-byte. Check knitbrain context meter periodically; when it says to, save a handoff with knitbrain save handoff and start a fresh session knitbrain load session restores everything . When the user states a task, call knitbrain run first and follow its directive skill + agents + commands . 证据：`.claude/rules/knitbrain.md`
- **Terse mode output tokens**（documentation）：Knit Brain compresses large tool outputs into skeletons. A ⟨ccr:HASH⟩ marker means the exact original is stored locally — call the knitbrain retrieve tool with that hash to read it byte-for-byte. Check knitbrain context meter periodically; when it says to, save a handoff with knitbrain save handoff and start a fresh session knitbrain load session restores everything . When the user states a task, call knitbrain run first and follow its directive skill + agents + commands . 证据：`.windsurf/rules/knitbrain.md`
- **Settings**（structured_config）：{ "hooks": { "PreToolUse": { "matcher": "Read", "hooks": { "type": "command", "command": "knitbrain-hook pretooluse" } } , "PreCompact": { "matcher": "", "hooks": { "type": "command", "command": "knitbrain-hook precompact" } } } } 证据：`.claude/settings.json`
- **.Mcp**（structured_config）：{ "mcpServers": { "knitbrain": { "command": "knitbrain" } } } 证据：`.mcp.json`
- **Tsconfig**（structured_config）：{ "compilerOptions": { "target": "ES2022", "module": "NodeNext", "moduleResolution": "NodeNext", "lib": "ES2022" , "types": "node" , "outDir": "dist", "rootDir": ".", "declaration": true, "sourceMap": true, "strict": true, "noUncheckedIndexedAccess": true, "noImplicitOverride": true, "noFallthroughCasesInSwitch": true, "verbatimModuleSyntax": true, "skipLibCheck": true, "esModuleInterop": true, "forceConsistentCasingInFileNames": true }, "include": "src/ / .ts", "tests/ / .ts", "scripts/ / .ts" , "exclude": "dist", "node modules" } 证据：`tsconfig.json`
- **transient / local-only**（source_file）：node modules/ dist/ coverage/ .log .DS Store .env .env. experiments.tsv 证据：`.gitignore`
- **Consistency**（source_file）：/ Consistency gate — kills the staleness problem structurally. Derives the truth from the CODE and fails the build when any doc/script hardcodes a drifted number. Runs as part of npm run verify after build . / ⋮---- const ok = cond, msg = ⋮---- // Source of truth: the built code. ⋮---- // Every "N tools" mention in README must equal the real count. ⋮---- // The production audit must not assert a stale hardcoded count. ⋮---- // Bins declared must exist in dist after build. 证据：`scripts/consistency.mjs`
- **E2E Tools**（source_file）：/ DEEP per-tool E2E against the BUILT artifact: spawn dist/index.js as a real MCP server over stdio isolated HOME so no developer state leaks in , then exercise EVERY tool with realistic arguments and assert on the substance of each response — the closed loop end to end: handshake instructions → load session auto knowledge init → run → classify plan-mode directive → FP loop → optimize/retrieve lossless → read → meter savings counted → learnings → handoff → resume → skills → agents → knowledge queries → team board lifecycle → metrics → ping. Run after npm run build . Exits non-zero on any failure. / ⋮---- const ok = cond, msg = ⋮---- function makeClient proc ⋮---- const rpc = method, params… 证据：`scripts/e2e-tools.mjs`
- **E2E**（source_file）：/ End-to-end check of the BUILT artifact dist/ : 1. spawn the MCP server, drive a real stdio session: initialize → tools/list → ping → optimize → retrieve the reverse loop , asserting a lossless round-trip over the wire. 2. run the unified compress on REAL files, asserting never-expand + byte-for-byte CCR recovery, reporting real token savings. Run after npm run build . Exits non-zero on any failure. / ⋮---- const distUrl = p ⋮---- const ok = cond, msg = ⋮---- function makeClient proc ⋮---- const rpc = method, params const notify = method, params ⋮---- const text = resp ⋮---- async function mcpSession ⋮---- // Memory loop rung 8 : record → load session reflects it. ⋮---- // Workflow + knowl… 证据：`scripts/e2e.mjs`
- **Production Audit**（source_file）：/ PRODUCTION AUDIT — cold-start, portable proof. Simulates shipping knitbrain to a different machine: Stage 1 fresh git clone of the committed state into a temp dir Stage 2 clean npm ci install there Stage 3 all 5 gates typecheck, lint, test, build, bench + e2e in the clone Stage 4 npm pack → install the tarball into a brand-new consumer project exactly what npm i knitbrain would deliver Stage 5 run the INSTALLED knitbrain binary: drive a full MCP session over stdio — exercise EVERY tool 20/20 with assertions Stage 6 run the INSTALLED knitbrain-proxy binary: real HTTP loop against a local fake upstream compression on the wire + SSE passthrough Stage 7 knitbrain setup CLI in the consumer pro… 证据：`scripts/production-audit.mjs`
- **Research Measure**（source_file）：/ Research worker — ONE measurement in a FRESH process, so WASM/heap state never accumulates across experiments the autoresearch-faithful design: each experiment is isolated . Parameter overrides arrive via the KNITBRAIN env vars that src/optimizer/params.ts already reads at load. Prints a single JSON line: {savings, pass, blocks}. Run by scripts/research.mjs. / ⋮---- const distUrl = p ⋮---- const noop = = 证据：`scripts/research-measure.mjs`
- **Research**（source_file）：/ knitbrain research — an autoresearch-style autonomous tuning loop for knitbrain's OWN compression heuristics inspired by karpathy/autoresearch . The product optimizes the user's agent loop; this optimizes the product. Hand-tuned constants anchor trigger, min-sentences, never-expand floor … are swept against REAL transcripts. The objective is the same hard number a user sees — overall savings % from profile — under a HARD CONSTRAINT: the fidelity gates evals must still pass. A setting that saves more but breaks an answer is auto-discarded, exactly like a crashed experiment in autoresearch. No mocks: every measurement runs on your own ~/.claude transcripts. Coordinate descent: from the curr… 证据：`scripts/research.mjs`
- **Shape Profile**（source_file）：/ Shape profiler — answers "where does context burn ACTUALLY live?" with data. Scans real host transcripts Claude Code JSONL , buckets every sizable tool result by shape, and reports per shape: count, tokens, and what the CURRENT optimizer already saves. The under-served heavy buckets are the next handlers to build — measurement decides, not guesswork. Usage: node scripts/shape-profile.mjs more... defaults to ~/.claude/projects — all projects, all sessions / ⋮---- const d = p ⋮---- // ── shape classification deterministic, order matters ── function dupRatio lines function classifyShape t ⋮---- try { JSON.parse t ; return "json"; } catch { / fallthrough / } ⋮---- // ── collect transcripts ──… 证据：`scripts/shape-profile.mjs`
- **Atomic**（source_file）：import { renameSync, writeFileSync } from "node:fs"; export function writeAtomic path: string, data: string Buffer, opts?: 证据：`src/atomic.ts`
- **Compress File**（source_file）：import { existsSync, lstatSync, readFileSync } from "node:fs"; import { countTokens } from "./tokenizer.js"; import { writeAtomic } from "./atomic.js"; ⋮---- / ^ \n + /g, // inline code /\bhttps?:\/\/\S+/gi, // URLs /\b \w.- /\\ \w./\\- +/g, // filesystem paths have a / or \ ⋮---- // Dropped from prose only never inside protected segments . ⋮---- // Sentinel = NUL char; cannot collide with real text incl. bare numbers . ⋮---- / Terse-rewrite prose; protected segments survive byte-for-byte. / export function compressProse text: string : string ⋮---- .replace / + .,;:!? /g, "$1" // tidy space left before punctuation ⋮---- / CLI: compress a file in place, keeping a verbatim .original backup. /… 证据：`src/compress-file.ts`
- **Dashboard**（source_file）：import { createServer, type Server } from "node:http"; import { fetchHubBoard, loadHubConfig } from "./hub/client.js"; import type { CCRStore } from "./ccr/store.js"; import type { Knowledge } from "./engine/knowledge.js"; import type { Memory } from "./engine/memory.js"; import type { Feedback } from "./engine/feedback.js"; import type { SkillsStore } from "./engine/skills.js"; import type { TeamBoard } from "./engine/teams.js"; import type { Meter } from "./engine/meter.js"; import type { PlatformUsage } from "./engine/usage.js"; import type { PlatformQuota } from "./engine/quota.js"; import type { ActivityEvent, AgentRollup } from "./engine/activity.js"; export interface DashboardDeps {… 证据：`src/dashboard.ts`
- **Evals**（source_file）：import { mkdtempSync, rmSync } from "node:fs"; import { homedir, tmpdir } from "node:os"; import { join } from "node:path"; import { createReadStream, readdirSync, statSync } from "node:fs"; import { createInterface } from "node:readline"; import { createFileCCRStore } from "./ccr/store.js"; import { ensureAst, astReady } from "./optimizer/ast.js"; import { compress, stripLineNumbers } from "./optimizer/router.js"; import { IMPORTANT LINE } from "./optimizer/structured.js"; export interface EvalReport { blocks: number; errorLines: { total: number; preserved: number }; identifiers: { total: number; preserved: number }; summaryLines: { total: number; preserved: number }; roundTrip: { total: n… 证据：`src/evals.ts`
- **Fan**（source_file）：import { spawn, spawnSync } from "node:child process"; import { existsSync, mkdirSync, readFileSync, writeFileSync } from "node:fs"; import { join } from "node:path"; export function parseTasks text: string : string export function markDone text: string, task: string : string interface FanOpts { goalFile?: string; workers: number; agent: string; verify: string null; max: number; isolate: boolean; } function parseArgs args: string : FanOpts function run cmd: string, cwd: string, input?: string : Promise function worktreeFor id: number, isolate: boolean : string function spawnSyncQuiet cmd: string : boolean function buildPrompt task: string : string export async function runFan args: string :… 证据：`src/fan.ts`
- **Global Config**（source_file）：import { existsSync, mkdirSync, readFileSync } from "node:fs"; import { writeAtomic } from "./atomic.js"; import { dirname, join } from "node:path"; export type GlobalConfigKind = "codex" "windsurf" "zed" "copilot-cli"; ⋮---- export function mergeGlobalConfig kind: GlobalConfigKind, existing: string null, : ⋮---- // JSON kinds: parse treat malformed as empty — it's backed up first , merge. ⋮---- export interface GlobalConfigIO { exists: p: string = boolean; read: p: string = string; write: p: string, data: string = void; mkdirp: dir: string = void; } ⋮---- export type GlobalConfigStatus = "written" "present" "created"; export function applyGlobalConfig kind: GlobalConfigKind, home: string,… 证据：`src/global-config.ts`
- **Index**（source_file）：import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js"; import { buildServer } from "./server.js"; import { runSetup } from "./setup.js"; function fmtTokens n: number : string async function main : Promise 证据：`src/index.ts`
- **Learn**（source_file）：import { createReadStream, existsSync, readdirSync, readFileSync, statSync, writeFileSync } from "node:fs"; import { homedir } from "node:os"; import { basename, join, resolve } from "node:path"; import { createInterface } from "node:readline"; interface ToolEvent { tool: string; input: Record ; error: boolean; errText: string; seq: number; } export interface Learning { category: "paths" "commands" "environment" "large-files"; text: string; count: number; } ⋮---- function publishable text: string : boolean ⋮---- export async function parseSession file: string : Promise ⋮---- // Trust the explicit flag; the regex is only a fallback when the // transcript doesn't carry one ordinary output oft… 证据：`src/learn.ts`
- **Lib**（source_file）：import { createFileCCRStore, type CCRStore } from "./ccr/store.js"; import { compress as routeCompress, type RouteResult, type CompressOptions } from "./optimizer/router.js"; import { ensureAst } from "./optimizer/ast.js"; import { ccrRoot } from "./paths.js"; import { countTokens } from "./tokenizer.js"; export interface Optimizer { compress text: string, options?: CompressOptions : RouteResult; retrieve handle: string : string; has handle: string : boolean; ready : Promise ; } ⋮---- compress text: string, options?: CompressOptions : RouteResult; retrieve handle: string : string; has handle: string : boolean; ready : Promise ; ⋮---- export interface OptimizerOptions { root?: string; } expo… 证据：`src/lib.ts`
- **Loop**（source_file）：import { spawnSync } from "node:child process"; import { existsSync, readFileSync, writeFileSync, appendFileSync } from "node:fs"; import { createInterface } from "node:readline/promises"; ⋮---- interface LoopOpts { goalFile?: string; max: number; agent: string; verify: string null; interactive: boolean; } function parseArgs args: string : LoopOpts function buildPrompt task: string, progress: string : string function run cmd: string, input?: string : boolean export async function runLoop args: string : Promise ⋮---- // Re-read before marking: the agent may have edited the goal file, and // writing the pre-agent text back would clobber those edits. 证据：`src/loop.ts`
- **Measure**（source_file）：import { countTokens } from "./tokenizer.js"; export interface Measurement { readonly label: string; readonly originalTokens: number; readonly optimizedTokens: number; readonly ratio: number; readonly savedPct: number; } export function measure label: string, original: string, optimized: string, : Measurement export function summarize measurements: readonly Measurement : 证据：`src/measure.ts`
- **Paths**（source_file）：import { createHash } from "node:crypto"; import { homedir } from "node:os"; import { join } from "node:path"; export function knitbrainHome : string export function ccrRoot : string export function projectId : string export function memoryRoot : string export function knowledgeRoot : string export function feedbackRoot : string export function teamRoot : string export function meterRoot : string export function skillsRoot : string export function calibrationRoot : string export function activityRoot : string 证据：`src/paths.ts`
- **Platforms**（source_file）：import { existsSync, mkdirSync, readFileSync } from "node:fs"; import { writeAtomic } from "./atomic.js"; import { dirname, join } from "node:path"; import type { SetupConfig } from "./setup.js"; export interface Artifact { path: string; content: string; mode: "write" "write-if-absent" "json-merge-mcp" "json-merge-hooks"; } ⋮---- export type TerseLevel = "lite" "full" "ultra"; export function terseGuide level: TerseLevel = "full" : string export function universalArtifacts : Artifact export function claudeArtifacts cfg: SetupConfig : Artifact export function cursorArtifacts : Artifact export function copilotSnippet : string export function vscodeArtifacts : Artifact export function windsurf… 证据：`src/platforms.ts`
- **Profile**（source_file）：import { createReadStream, mkdtempSync, readdirSync, rmSync, statSync } from "node:fs"; import { createInterface } from "node:readline"; import { homedir, tmpdir } from "node:os"; import { join } from "node:path"; import { sha256, createFileCCRStore } from "./ccr/store.js"; import { compress } from "./optimizer/router.js"; import { ensureAst, astReady } from "./optimizer/ast.js"; import { countTokens } from "./tokenizer.js"; function dupRatio lines: string : number export function classifyShape t: string : string function collectTranscripts roots: string : string interface Bucket { n: number; before: number; after: number; } export async function runProfile args: string , log: line: string… 证据：`src/profile.ts`
- **Server**（source_file）：import { Server } from "@modelcontextprotocol/sdk/server/index.js"; import { CallToolRequestSchema, ListToolsRequestSchema, } from "@modelcontextprotocol/sdk/types.js"; import { createFileCCRStore, type CCRStore } from "./ccr/store.js"; import { createMemory, type Memory } from "./engine/memory.js"; import { createKnowledge, type Knowledge } from "./engine/knowledge.js"; import { createFeedback, type Feedback } from "./engine/feedback.js"; import { createTeamBoard, type TeamBoard } from "./engine/teams.js"; import { createMeter, type Meter } from "./engine/meter.js"; import { createSkillsStore, type SkillsStore } from "./engine/skills.js"; import { createCalibration, type Calibration } from… 证据：`src/server.ts`
- **Tokenizer**（source_file）：import { encode } from "gpt-tokenizer/encoding/o200k base"; export interface Tokenizer { readonly name: string; count text: string : number; } ⋮---- count text: string : number; ⋮---- count text: string : number ⋮---- export function setTokenizer tokenizer: Tokenizer : void export function countTokens text: string : number export function activeTokenizerName : string 证据：`src/tokenizer.ts`
- **Wrap**（source_file）：import { spawn } from "node:child process"; import { dirname, join } from "node:path"; import { fileURLToPath } from "node:url"; export interface WrapPlan { binary: string; envVar: "ANTHROPIC BASE URL" "OPENAI BASE URL"; baseUrl: string; } ⋮---- export function resolveWrap agent: string, port: number = DEFAULT PROXY PORT : WrapPlan export function hasApiKey env: NodeJS.ProcessEnv : boolean async function proxyHealthy port: number : Promise function proxyEntry : string export async function runWrap argv: string : Promise 证据：`src/wrap.ts`

## 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：`CLAUDE.md`, `README.md`, `package.json`
- **回答用户时区分可预览内容与必须安装后才能验证的内容。**：安装前体验的消费者价值来自降低误装和误判，而不是伪装成真实运行。 证据：`CLAUDE.md`, `README.md`, `package.json`

## 用户开工前应该回答的问题

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

## 验收标准

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

---

## Doramagic Context Augmentation

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

## Human Manual 骨架

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

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

- **项目概述与系统架构**：importance `high`
  - source_paths: README.md, package.json, src/index.ts, src/lib.ts, src/server.ts
- **无损上下文压缩与 Read 优化**：importance `high`
  - source_paths: src/optimizer/ast.ts, src/optimizer/code.ts, src/optimizer/json.ts, src/optimizer/params.ts, src/optimizer/router.ts
- **记忆、知识图谱与层级路由工作流**：importance `high`
  - source_paths: src/engine/memory.ts, src/engine/knowledge.ts, src/engine/calibration.ts, src/engine/feedback.ts, src/engine/meter.ts
- **自治循环、多平台集成与运维**：importance `medium`
  - source_paths: src/loop.ts, src/fan.ts, src/hooks/index.ts, src/hooks/pretooluse.ts, src/hooks/sessionstart.ts

## Repo Inspection Evidence / 源码检查证据

- repo_clone_verified: true
- repo_inspection_verified: true
- repo_commit: `b5b0bffcd92d7040aae9c51b7582aeea0915e8f8`
- inspected_files: `README.md`, `package.json`, `docs/ROADMAP.md`, `src/atomic.ts`, `src/ccr/store.ts`, `src/compress-file.ts`, `src/dashboard.ts`, `src/engine/activity.ts`, `src/engine/agents.ts`, `src/engine/calibration.ts`, `src/engine/feedback.ts`, `src/engine/knowledge.ts`, `src/engine/memory.ts`, `src/engine/meter.ts`, `src/engine/quota.ts`, `src/engine/skills.ts`, `src/engine/teams.ts`, `src/engine/usage.ts`, `src/engine/workflow.ts`, `src/evals.ts`

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/PDgit12/knitbrain | 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/PDgit12/knitbrain | 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: 运行可能依赖外部服务

- Trigger: 项目说明出现 external service/cloud/webhook/database 等运行依赖关键词。
- Host AI rule: 确认是否有离线 demo、mock 数据或可替代服务。
- Why it matters: 本地安装成功不等于能力可用，外部服务不可用会阻断体验。
- Evidence: packet_text.keyword_scan | https://github.com/PDgit12/knitbrain | matched external service / cloud / webhook / database keyword
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 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 | https://github.com/PDgit12/knitbrain | 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/PDgit12/knitbrain | 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
- 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/PDgit12/knitbrain | 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 | https://github.com/PDgit12/knitbrain | 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 | https://github.com/PDgit12/knitbrain | 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。