# forgekit - 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 forgekit 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

- **AI 研究者或研究型 Agent 构建者**：README 明确围绕研究、实验或论文工作流展开。 证据：`README.md` Claim：`clm_0004` supported 0.86
- **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_0005` 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：`global/tools/atlas/SKILL.md`, `global/tools/catchup/SKILL.md`, `global/tools/code-modernization/SKILL.md`, `global/tools/cognitive-substrate/SKILL.md` 等 Claim：`clm_0006` 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：`global/tools/atlas/SKILL.md`, `global/tools/catchup/SKILL.md`, `global/tools/code-modernization/SKILL.md`, `global/tools/cognitive-substrate/SKILL.md` 等 Claim：`clm_0001` supported 0.86
- **多宿主安装与分发**（需要安装后验证）：项目包含插件或 marketplace 配置，说明它面向一个或多个 AI 宿主的安装和分发。 证据：`.claude-plugin/marketplace.json`, `.claude-plugin/plugin.json`, `.codex-plugin/plugin.json` Claim：`clm_0002` supported 0.86
- **命令行启动或安装流程**（需要安装后验证）：项目文档中存在可执行命令，真实使用需要在本地或宿主环境中运行这些命令。 证据：`README.md` Claim：`clm_0003` supported 0.86

## 怎么开始

- `npm install -g @codewithjuber/forgekit   # or: npm install -g github:CodeWithJuber/forgekit` 证据：`README.md` Claim：`clm_0007` 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：AI 研究者或研究型 Agent 构建者**（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_0004` supported 0.86
- **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_0005` 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：`global/tools/atlas/SKILL.md`, `global/tools/catchup/SKILL.md`, `global/tools/code-modernization/SKILL.md`, `global/tools/cognitive-substrate/SKILL.md` 等 Claim：`clm_0006` 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：`global/tools/atlas/SKILL.md`, `global/tools/catchup/SKILL.md`, `global/tools/code-modernization/SKILL.md`, `global/tools/cognitive-substrate/SKILL.md` 等 Claim：`clm_0001` supported 0.86
- **能力存在：多宿主安装与分发**（supported）：可以相信项目包含这类能力线索；是否适合你的具体任务仍要试用或安装后验证。 证据：`.claude-plugin/marketplace.json`, `.claude-plugin/plugin.json`, `.codex-plugin/plugin.json` Claim：`clm_0002` supported 0.86
- **能力存在：命令行启动或安装流程**（supported）：可以相信项目包含这类能力线索；是否适合你的具体任务仍要试用或安装后验证。 证据：`README.md` Claim：`clm_0003` supported 0.86

### 现在还不能相信

- **研究结论、引用和实验结果不能在安装前相信。**（unverified）：研究 Skill 可以组织问题和路径，但不能替代真实资料检索、论文核验和实验复现。
- **是否适合你的具体研究领域不能直接相信。**（unverified）：Skill 覆盖很多研究主题，不代表对你的领域、资料要求和可信度标准足够。
- **真实输出质量不能在安装前相信。**（unverified）：Prompt Preview 只能展示引导方式，不能证明真实项目中的结果质量。
- **宿主 AI 版本兼容性不能在安装前相信。**（unverified）：Claude、Cursor、Codex、Gemini 等宿主加载规则和版本差异必须在真实环境验证。
- **不会污染现有宿主 AI 行为，不能直接相信。**（inferred）：Skill、plugin、AGENTS/CLAUDE/GEMINI 指令可能改变宿主 AI 的默认行为。 证据：`.claude-plugin/marketplace.json`, `.claude-plugin/plugin.json`, `.codex-plugin/plugin.json`, `CLAUDE.md` 等
- **可安全回滚不能默认相信。**（unverified）：除非项目明确提供卸载和恢复说明，否则必须先在隔离环境验证。
- **真实安装后是否与用户当前宿主 AI 版本兼容？**（unverified）：兼容性只能通过实际宿主环境验证。 证据：`.claude-plugin/marketplace.json`, `.claude-plugin/plugin.json`, `.codex-plugin/plugin.json`
- **项目输出质量是否满足用户具体任务？**（unverified）：安装前预览只能展示流程和边界，不能替代真实评测。

### 继续会触碰什么

- **研究判断**：问题拆解、资料路径、实验路径、结论结构和可信度判断。 原因：研究型 Skill 可能让输出看起来更专业，但不能替代真实证据核验。
- **命令执行**：包管理器、网络下载、本地插件目录、项目配置或用户主目录。 原因：运行第一条命令就可能产生环境改动；必须先判断是否值得跑。 证据：`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-plugin/marketplace.json`, `.claude-plugin/plugin.json`, `.codex-plugin/plugin.json`, `CLAUDE.md` 等
- **本地环境或项目文件**：安装结果、插件缓存、项目配置或本地依赖目录。 原因：安装前无法证明写入范围和回滚方式，需要隔离验证。 证据：`.claude-plugin/marketplace.json`, `.claude-plugin/plugin.json`, `.codex-plugin/plugin.json`, `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**：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_0008` inferred 0.45
- **宿主 AI 插件或 Skill 规则冲突**：新规则可能改变用户现有宿主 AI 的工作方式。 处理方式：安装前先检查插件 manifest 和 Skill 文件，必要时隔离测试。 证据：`.claude-plugin/marketplace.json`, `.claude-plugin/plugin.json`, `.codex-plugin/plugin.json` Claim：`clm_0009` supported 0.86
- **命令执行会修改本地环境**：安装命令可能写入用户主目录、宿主插件目录或项目配置。 处理方式：先在隔离环境或测试账号中运行。 证据：`README.md` Claim：`clm_0010` 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：`global/tools/atlas/SKILL.md`, `global/tools/catchup/SKILL.md`, `global/tools/code-modernization/SKILL.md`, `global/tools/cognitive-substrate/SKILL.md` 等 Claim：`clm_0001` supported 0.86
- **多宿主安装与分发**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`.claude-plugin/marketplace.json`, `.claude-plugin/plugin.json`, `.codex-plugin/plugin.json` Claim：`clm_0002` supported 0.86
- **命令行启动或安装流程**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`README.md` Claim：`clm_0003` supported 0.86

### 上下文规模

- 文件总数：464
- 重要文件覆盖：40/464
- 证据索引条目：80
- 角色 / Skill 条目：19

### 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 forgekit 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 forgekit 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 forgekit 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 19 个角色 / Skill / 项目文档条目。

- **atlas**（skill）：Forge's code-graph. Use to find where a symbol is defined where-is-X , to discover reusable code before writing new, and to check a symbol actually exists before calling it anti-hallucination . Backed by forge atlas . 激活提示：当用户任务与“atlas”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：`global/tools/atlas/SKILL.md`
- **catchup**（skill）：Start-of-session context rebuild. Use when resuming work in a repo after a gap, a /clear, or a machine switch — "where were we?" answered from recorded state goal, session snapshot, decisions, recent commits instead of assumptions. 激活提示：当用户任务与“catchup”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：`global/tools/catchup/SKILL.md`
- **code-modernizer**（skill）：- 激活提示：当用户任务与“code-modernizer”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：`global/tools/code-modernization/SKILL.md`
- **cognitive-substrate**（skill）：- 激活提示：当用户任务与“cognitive-substrate”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：`global/tools/cognitive-substrate/SKILL.md`
- **cost-guard**（skill）：Reduce token/cost burn on large or exploratory tasks. Use when a task will touch many files, involves broad codebase search, a big migration, or when the user asks to keep cost/context low. 激活提示：当用户任务与“cost-guard”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：`global/tools/cost-guard/SKILL.md`
- **design-md**（skill）：Create and maintain a per-project DESIGN.md that pins the ONE visual direction so UI stays consistent across sessions and screens. Use when starting UI on a project, when look feels inconsistent, or the user mentions design system / DESIGN.md / visual guidelines. 激活提示：当用户任务与“design-md”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：`global/tools/design-md/SKILL.md`
- **dev-radar**（skill）：Pull current, real-world software-dev signal on demand from GitHub trending, Reddit, research papers, and blogs — filtered to the user's stack, hype-filtered, with verified sources. Use when the user asks "what's new/trending", "latest in ", "should I adopt X", or wants a dev digest. Coding topics only. 激活提示：当用户任务与“dev-radar”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：`global/tools/dev-radar/SKILL.md`
- **explore-plan-code**（skill）：Structured workflow for non-trivial features or unfamiliar code — explore, then plan, then implement, then verify. Use when a change spans multiple files, the approach is uncertain, or the user asks for a spec/plan first. 激活提示：当用户任务与“explore-plan-code”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：`global/tools/explore-plan-code/SKILL.md`
- **handoff**（skill）：End-of-session checkpoint. Use when finishing, pausing, or switching away from work in a repo — it rewrites the bounded session snapshot .forge/state.md the NEXT session is re-injected with, so nothing essential dies with this conversation. Backed by forge handoff . 激活提示：当用户任务与“handoff”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：`global/tools/handoff/SKILL.md`
- **lean**（skill）：Forge's minimalism discipline. Use for any "add X" / "build Y" / refactor / bug-fix — before writing code, to choose the smallest change that actually solves the problem. Reuse over rewrite, delete over add, boring over clever. 激活提示：当用户任务与“lean”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：`global/tools/lean/SKILL.md`
- **problem-solver**（skill）：Universal problem-solving engine that fuses proven frameworks ASQ 4-step cycle, DMAIC/PDCA, 5 Whys, Fishbone, First Principles, TRIZ, Cynefin, Design Thinking, weighted decision matrix into one staged loop. Use when the user asks to solve, analyze, or break down a problem, find root causes, choose between options, innovate, make a hard decision, handle a crisis, or wants structured/step-by-step problem solving. 激活提示：当用户任务与“problem-solver”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：`global/tools/problem-solver/SKILL.md`
- **recall**（skill）：Forge's cross-session memory. Use when the user says "remember this", when you learn a durable non-obvious fact/decision/gotcha worth keeping, or to recall past context. Backed by forge recall + the recall-load guard. 激活提示：当用户任务与“recall”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：`global/tools/recall/SKILL.md`
- **reuse-first**（skill）：Engineering standards for writing production code — reuse before building, follow existing patterns, testable/reusable design, and verify. Use for any non-trivial feature, refactor, or "add X" task, and when choosing what to build vs. adopt. 激活提示：当用户任务与“reuse-first”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：`global/tools/reuse-first/SKILL.md`
- **self-improve**（skill）：Self-correction and cross-session learning loop. Use when a check fails and needs iterating to green, when the user corrects the same thing more than once, or when a non-obvious fix/gotcha is worth remembering for next time. 激活提示：当用户任务与“self-improve”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：`global/tools/self-improve/SKILL.md`
- **sync-docs**（skill）：After ANY code change, sweep the diff for documentation that just went stale. Use when finishing an edit, before a commit/PR, or when the completion gate blocks — it turns "did I forget a doc?" into a mechanical checklist. Backed by forge docs sync . 激活提示：当用户任务与“sync-docs”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：`global/tools/sync-docs/SKILL.md`
- **taste**（skill）：Pick a visual design direction for a repo from Forge's menu minimalist, brutalist, editorial, playful, corporate . Use when starting UI work or when the user wants a specific look. Sets DESIGN.md, which every AI tool then follows. 激活提示：当用户任务与“taste”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：`global/tools/taste/SKILL.md`
- **tech-selector**（skill）：Choose a library, framework, or tool by verifying the CURRENT best option from live sources — not from training data. Use whenever picking a dependency, starting a project, or the user asks "what's the best X for Y", "latest", "which library", or wants an unbiased/current recommendation. 激活提示：当用户任务与“tech-selector”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：`global/tools/tech-selector/SKILL.md`
- **ui-workflow**（skill）：Reliable workflow for building or fixing UI so results look intentional, not generic. Use for any frontend/UI task — new screen, component, redesign, "make it look better", layout, styling, or when the user says UI is the pain point. 激活提示：当用户任务与“ui-workflow”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：`global/tools/ui-workflow/SKILL.md`
- **hostlelo-deploy**（skill）：Safe, staged deploy of Hostlelo services to the VPS. Use when asked to deploy, ship, or release a Hostlelo change kernel, widget, or Telegram control plane . 激活提示：当用户任务与“hostlelo-deploy”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：`templates/project-layer/.claude/skills/hostlelo-deploy/SKILL.md`

## 证据索引

- 共索引 80 条证据。

- **The Forge Cognitive Substrate**（documentation）：Coding agents forget what they learned, assume what they don't know, and break code they can't see. The substrate is a fast, mostly-deterministic check that runs before an agent edits your code: it flags an unclear task, picks the cheapest capable model, and shows what an edit will break — all from the repo you already have, with no extra LLM call. 证据：`docs/cognitive-substrate/README.md`
- **Forge — one shared brain for your AI coding tools**（documentation）：Forge — one shared brain for your AI coding tools 证据：`README.md`
- **Examples**（documentation）：Short, copy-pasteable walkthroughs. forgekit is one brain for every AI coding agent — the cognitive substrate memory, foresight, guardrails that a stateless model is missing. These examples show that brain doing its job: gating an edit before it happens, and carrying what your team learns from one machine to the next. 证据：`examples/README.md`
- **Forge documentation site Mintlify**（documentation）：This folder is a self-contained Mintlify https://mintlify.com documentation site for Forge @codewithjuber/forgekit . It is additive — it does not replace the Markdown docs in docs/ , README.md , ARCHITECTURE.md , or ONBOARDING.md , which remain the source of truth. The site content is derived from those files. 证据：`mintlify/README.md`
- **Formal Synthesis — A Theory of the Cognitive Substrate for Coding Agents**（documentation）：Formal Synthesis — A Theory of the Cognitive Substrate for Coding Agents 证据：`research/formal-synthesis/README.md`
- **Python prototypes**（documentation）：These are the original evidence prototypes from the cognitive-substrate paper. They are preserved for auditability and result reproduction only. 证据：`research/python-prototypes/README.md`
- **Impact Oracle — Codebase World-Model + Blast-Radius Predictor**（documentation）：Impact Oracle — Codebase World-Model + Blast-Radius Predictor 证据：`research/python-prototypes/impact_oracle/README.md`
- **router gate**（documentation）：A research prototype of an assumption gate + complexity-aware model router for coding agents the auditable Python original of forge preflight / forge route ; a demonstration, not a benchmark . 证据：`research/python-prototypes/router_gate/README.md`
- **Package**（package_manifest）：{ "name": "@codewithjuber/forgekit", "version": "0.27.3", "description": "Shared memory, impact analysis, and guardrail hooks for AI coding agents — authored once, emitted as native config for Claude Code, Codex, Cursor, Gemini, Aider, and more.", "type": "module", "bin": { "forge": "src/cli.js" }, "engines": { "node": " =20" }, "exports": { ".": "./src/cli.js", "./model-tiers": "./src/model tiers.js", "./package.json": "./package.json" }, "author": "CodeWithJuber", "license": "MIT", "repository": { "type": "git", "url": "git+https://github.com/CodeWithJuber/forgekit.git" }, "bugs": { "url": "https://github.com/CodeWithJuber/forgekit/issues" }, "homepage": "https://github.com/CodeWithJuber/… 证据：`package.json`
- **forgekit — contributor instructions**（documentation）：forgekit — contributor instructions 证据：`CLAUDE.md`
- **Global instructions all projects**（documentation）：Loaded every session. Keep short — long files get ignored. Prune ruthlessly. 证据：`global/CLAUDE.md`
- **AGENTS.md —**（documentation）：Cross-tool rules read by Codex, Cursor, Copilot, Gemini, Aider, Zed, and by Claude Code when CLAUDE.md points here . Keep tool-agnostic and thin. 证据：`templates/project-layer/AGENTS.md`
- **— project instructions**（documentation）：Drop this into a repo root. Adds project context on top of the global ~/.claude/CLAUDE.md . Keep it lean; prune anything Claude gets right without it. 证据：`templates/project-layer/CLAUDE.md`
- **Contributing to forgekit**（documentation）：forgekit is one brain for every AI coding agent — a cognitive substrate memory, foresight, guardrails authored once and compiled into every tool's native config. Keeping that brain small and dependency-free is the whole point, so please read this before opening a PR. 证据：`CONTRIBUTING.md`
- **atlas — the code-graph**（skill_instruction）：A precomputed, portable symbol index at .forge/atlas.json . Any tool reads it via forge atlas or plain jq — no MCP required to consume. 证据：`global/tools/atlas/SKILL.md`
- **catchup — resume, don't re-assume**（skill_instruction）：A fresh session that guesses the project's state re-burns every iteration the last one already paid for. Everything essential is recorded; read it BEFORE acting. 证据：`global/tools/catchup/SKILL.md`
- **Code Modernizer**（skill_instruction）：A senior engineer isn't measured by how much code they wrote today, but by how little the system needs. Before touching anything, decide whether the code should be replaced with something already established. If custom code is genuinely necessary, write the least of it that solves the real problem without dropping behavior anyone relies on. 证据：`global/tools/code-modernization/SKILL.md`
- **Cognitive Substrate**（skill_instruction）：Wrap the frozen model in a pre-action check: gate assumptions, route model effort, predict blast radius, decompose scope, surface past lessons, and plan verification — before editing. 证据：`global/tools/cognitive-substrate/SKILL.md`
- **Cost guard**（skill_instruction）：Playbook for high-performance, low-cost runs. The constraint is the context window: performance degrades as it fills, and tokens cost money. Spend context deliberately. 证据：`global/tools/cost-guard/SKILL.md`
- **DESIGN.md**（skill_instruction）：The fix for "UI comes out inconsistent" is a single written source of visual truth per project. ui-workflow says pick one direction — DESIGN.md is where that direction lives so every session and screen obeys the same rules. forge taste writes a managed one from Forge's menu; hand-write it only for a bespoke direction. 证据：`global/tools/design-md/SKILL.md`
- **Dev radar**（skill_instruction）：On-demand scan of what's actually happening in software dev right now — not training-data recall. Signal over noise, coding topics only. 证据：`global/tools/dev-radar/SKILL.md`
- **Explore → Plan → Code → Verify**（skill_instruction）：Separate research from execution so you don't build the wrong thing. 证据：`global/tools/explore-plan-code/SKILL.md`
- **handoff — persist what this session knows**（skill_instruction）：handoff — persist what this session knows 证据：`global/tools/handoff/SKILL.md`
- **lean — the smallest change that works**（skill_instruction）：lean — the smallest change that works 证据：`global/tools/lean/SKILL.md`
- **Problem Solver**（skill_instruction）：This skill solves any problem — technical, business, personal, or complex — through a staged loop. Each analytical stage carries a discipline that keeps the analysis honest e.g. verify facts before diagnosing root cause . Match depth to the problem: a trivial problem gets two sentences of method, not all six stages. 证据：`global/tools/problem-solver/SKILL.md`
- **recall — durable cross-session memory**（skill_instruction）：recall — durable cross-session memory 证据：`global/tools/recall/SKILL.md`
- **Reuse-first engineering**（skill_instruction）：Default to reusing and following what exists. New code is a liability; the best change is the smallest one that fits the codebase. 证据：`global/tools/reuse-first/SKILL.md`
- **Self-improve**（skill_instruction）：Honest scope: you can't reinforcement-learn a hosted model's weights in a config. What works is experiential learning — correct within the session, persist durable lessons for the next one. Two loops: 证据：`global/tools/self-improve/SKILL.md`
- **sync-docs — make every artifact true again**（skill_instruction）：sync-docs — make every artifact true again 证据：`global/tools/sync-docs/SKILL.md`
- **taste — choose one design direction**（skill_instruction）：taste — choose one design direction 证据：`global/tools/taste/SKILL.md`
- **Tech selector**（skill_instruction）：Training data is stale and biased toward what was popular at cutoff. For any "what should I use" decision, verify against live sources before recommending. 证据：`global/tools/tech-selector/SKILL.md`
- **UI workflow**（skill_instruction）：UI is unreliable when taste is vague and there's no check. Fix both: lock ONE visual direction, build on the project's real components, then verify with the deterministic design gate and a screenshot instead of guessing. 证据：`global/tools/ui-workflow/SKILL.md`
- **Hostlelo deploy**（skill_instruction）：Staged deploy with a check at each gate. Never skip verification. Never deploy with a failing build or uncommitted secrets. 证据：`templates/project-layer/.claude/skills/hostlelo-deploy/SKILL.md`
- **Marketplace**（structured_config）：{ "$schema": "https://json.schemastore.org/claude-code-marketplace.json", "name": "forge", "owner": { "name": "CodeWithJuber" }, "plugins": { "name": "forgekit", "source": ".", "description": "One config, every AI coding tool. Cross-tool rules emitter, enforced guards, code-graph, memory, and cost governor." } } 证据：`.claude-plugin/marketplace.json`
- **Plugin**（structured_config）：{ "$schema": "https://json.schemastore.org/claude-code-plugin-manifest.json", "name": "forgekit", "displayName": "Forge", "version": "0.27.3", "description": "One config, every AI coding tool — cognitive substrate, tools, crew, guards, atlas, lean, recall from one source.", "author": { "name": "CodeWithJuber" }, "license": "MIT", "repository": "https://github.com/CodeWithJuber/forgekit", "homepage": "https://github.com/CodeWithJuber/forgekit readme", "keywords": "config", "cross-tool", "agents-md", "ai-coding", "claude-code", "cognitive-substrate" , "skills": "./global/tools", "agents": "./global/crew/doc-sync.md", "./global/crew/frontend-verifier.md", "./global/crew/independent-reviewer.md… 证据：`.claude-plugin/plugin.json`
- **Plugin**（structured_config）：{ "name": "forgekit", "version": "0.27.3", "description": "One config, every AI coding tool — cognitive substrate, MCP tools, guards, atlas, recall, and routing from one source.", "author": { "name": "CodeWithJuber", "url": "https://github.com/CodeWithJuber" }, "homepage": "https://github.com/CodeWithJuber/forgekit readme", "repository": "https://github.com/CodeWithJuber/forgekit", "license": "MIT", "keywords": "codex", "mcp", "ai-coding", "agents-md", "cognitive-substrate" , "skills": "global/tools", "mcpServers": ".mcp.json", "interface": { "displayName": "Forge", "shortDescription": "Cognitive substrate and one config for every AI coding tool.", "longDescription": "Forge adds a pre-actio… 证据：`.codex-plugin/plugin.json`
- **License**（source_file）：Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files the "Software" , to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: 证据：`LICENSE`
- **forgekit — architecture**（documentation）：One brain for every AI coding agent. A large language model is stateless: one context window, wiped every call. It has no memory of what your team learned, no foresight about what an edit will break, and no enforced guardrails. forgekit is the cognitive substrate — the layer that runs before the model edits code, supplying proof-carrying memory, impact foresight, and enforced guardrails — and a cross-tool config compiler that delivers that brain as native config into every tool at once. 证据：`ARCHITECTURE.md`
- **Onboarding — five minutes to productive**（documentation）：Onboarding — five minutes to productive 证据：`ONBOARDING.md`
- **Forge — the complete guide**（documentation）：One brain for every AI coding agent. A language model is stateless — one context window, wiped every call. It can't remember what your team learned, can't foresee what an edit breaks, and has no enforced guardrails. Forge is the cognitive substrate — the layer that runs before the model edits code — that supplies exactly those three things, and it ships them as native config to nine AI coding tools at once. The brain is the point; one config for every tool is how the brain gets delivered. 证据：`docs/GUIDE.md`
- **Releasing forgekit**（documentation）：Releases are automatic. Merge to master and the release cuts itself — no button, no manual bump. Two workflows do everything: 证据：`docs/RELEASING.md`
- **Repository review notes**（documentation）：- Runtime dependency policy is consistent: package.json has no production dependencies and the docs describe a zero-runtime-dependency Node CLI. - Core user docs are present: README, onboarding, architecture, guide, changelog, governance, support, security, releasing, and accessibility files exist. - Benchmark claims in the README point readers to reports/benchmarks.md , which is the right direction for avoiding stale marketing numbers. 证据：`docs/REPO_REVIEW.md`
- **ADR NNNN:**（documentation）：- Status: proposed accepted superseded by ADR-XXXX - Date: YYYY-MM-DD 证据：`docs/adr/0000-template.md`
- **ADR 0001: Zero runtime dependencies**（documentation）：ADR 0001: Zero runtime dependencies 证据：`docs/adr/0001-zero-runtime-dependencies.md`
- **ADR 0002: Adopt the SKILL.md open standard; don't invent a format**（documentation）：ADR 0002: Adopt the SKILL.md open standard; don't invent a format 证据：`docs/adr/0002-adopt-skill-md-standard.md`
- **ADR 0003: Enforce invariants as guards; keep prose thin**（documentation）：ADR 0003: Enforce invariants as guards; keep prose thin 证据：`docs/adr/0003-guard-over-prose.md`
- **ADR 0004: Which 2026 production standards apply to a zero-dep CLI**（documentation）：ADR 0004: Which 2026 production standards apply to a zero-dep CLI 证据：`docs/adr/0004-standards-for-a-zero-dep-cli.md`
- **ADR 0005: Allow selective runtime dependencies**（documentation）：ADR 0005: Allow selective runtime dependencies 证据：`docs/adr/0005-allow-runtime-dependencies.md`
- **ADR 0006: Proof-Carrying Memory is the one store**（documentation）：ADR 0006: Proof-Carrying Memory is the one store 证据：`docs/adr/0006-proof-carrying-memory.md`
- **Audit remediation record**（documentation）：A concise record of the honesty / wording audit and what was addressed in this change. Status legend: Done shipped here · Deferred tracked, out of scope for this pass . 证据：`docs/audit/REMEDIATION.md`
- **A Cognitive Substrate for Coding Agents — Deliverable Package**（documentation）：A Cognitive Substrate for Coding Agents — Deliverable Package 证据：`docs/cognitive-substrate/deliverable-package.md`
- **Ecosystem Map: Cognitive Substrate vs. the Mid-2026 Claude Code Stack**（documentation）：Ecosystem Map: Cognitive Substrate vs. the Mid-2026 Claude Code Stack 证据：`docs/cognitive-substrate/ecosystem_map.md`
- **Evidence Map: Independent Re-Grounding of Pain-Point Statistics**（documentation）：Evidence Map: Independent Re-Grounding of Pain-Point Statistics 证据：`docs/cognitive-substrate/evidence_map.md`
- **Playbook — how to drive your Claude setup**（documentation）：Playbook — how to drive your Claude setup 证据：`docs/legacy/PLAYBOOK.md`
- **Audit of your live ~/.claude + how this bundle fits**（documentation）：Audit of your live ~/.claude + how this bundle fits 证据：`docs/legacy/RECONCILE.md`
- **Deep research — best Claude Code configs vs. yours 2026**（documentation）：Deep research — best Claude Code configs vs. yours 2026 证据：`docs/legacy/RESEARCH-claude-config.md`
- **RUN — command sheet current state**（documentation）：⚠️ Archived — pre-Forge. Kept for history; not maintained . It describes an older hand-rolled ~/.claude setup, not today's forge CLI. For the current workflow start at ONBOARDING.md ../../ONBOARDING.md → docs/GUIDE.md ../GUIDE.md . Any credentials or open action items below are historical and superseded — do not act on them. 证据：`docs/legacy/RUN.md`
- **Daily Config Radar — 2026-07-04**（documentation）：⚠️ Archived — pre-Forge. A historical auto-generated report, kept for the record. Not maintained and not the current state. See docs/GUIDE.md ../../GUIDE.md for today's tooling. 证据：`docs/legacy/reports/2026-07-04.md`
- **Substrate v2 — completing the whitepaper, and the Proof-Carrying Memory protocol**（documentation）：Substrate v2 — completing the whitepaper, and the Proof-Carrying Memory protocol 证据：`docs/plans/substrate-v2/00-overview.md`
- **01 — The Proof-Carrying Memory PCM protocol**（documentation）：01 — The Proof-Carrying Memory PCM protocol 证据：`docs/plans/substrate-v2/01-pcm-protocol.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/cognitive-substrate/README.md`, `README.md`, `examples/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/cognitive-substrate/README.md`, `README.md`, `examples/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。

- **Forge 概览、安装与快速入门**：importance `high`
  - source_paths: README.md, package.json, install.sh, bin/claude-init.sh, ARCHITECTURE.md
- **认知基座：行动前闸门（Substrate）**：importance `high`
  - source_paths: src/substrate.js, src/preflight.js, src/route.js, src/impact.js, src/predictor.js
- **证明性记忆（PCM）、账本与团队同步**：importance `high`
  - source_paths: src/ledger.js, src/ledger_store.js, src/ledger_sync.js, src/ledger_read.js, src/ledger_bridge.js
- **多工具分发、MCP、验证与可观测性**：importance `high`
  - source_paths: src/init.js, src/sync.js, src/integrations.js, src/doctor.js, src/update.js

## Repo Inspection Evidence / 源码检查证据

- repo_clone_verified: true
- repo_inspection_verified: true
- repo_commit: `766254eba2da15a658ac3f70b1f08f5227efe708`
- inspected_files: `README.md`, `package.json`, `docs/GUIDE.md`, `docs/RELEASING.md`, `docs/REPO_REVIEW.md`, `docs/adr/0000-template.md`, `docs/adr/0001-zero-runtime-dependencies.md`, `docs/adr/0002-adopt-skill-md-standard.md`, `docs/adr/0003-guard-over-prose.md`, `docs/adr/0004-standards-for-a-zero-dep-cli.md`, `docs/adr/0005-allow-runtime-dependencies.md`, `docs/adr/0006-proof-carrying-memory.md`, `docs/audit/REMEDIATION.md`, `docs/cognitive-substrate/README.md`, `docs/cognitive-substrate/deliverable-package.md`, `docs/cognitive-substrate/ecosystem_map.md`, `docs/cognitive-substrate/evidence_map.md`, `docs/legacy/PLAYBOOK.md`, `docs/legacy/RECONCILE.md`, `docs/legacy/RESEARCH-claude-config.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/CodeWithJuber/forgekit | host_targets=claude_code, claude, cursor, mcp_host
- 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/CodeWithJuber/forgekit | 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/CodeWithJuber/forgekit | 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/CodeWithJuber/forgekit | 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/CodeWithJuber/forgekit | 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/CodeWithJuber/forgekit | 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/CodeWithJuber/forgekit | 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。