bmad-speckit-sdd-flow 1.0.0 → 1.0.2

package/README.md

@@ -1,82 +1,137 @@
 # BMAD-Speckit-SDD-Flow
 
-[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](LICENSE)
-[![Node.js Version](https://img.shields.io/badge/node-%3E%3D18.0.0-brightgreen)](https://nodejs.org)
+English | [简体中文](README.zh-CN.md)
 
 <p align="center">
   <img src="docs/assets/readme-slogan.final.svg" alt="BMAD-Speckit-SDD-Flow slogan banner" width="100%" />
 </p>
 
-**Built on** [BMAD-METHOD](https://github.com/bmad-code-org/BMAD-METHOD) and [github/spec-kit](https://github.com/github/spec-kit).
-**Extended with** audit loops, critical auditor, scoring system, AI Coach, and SFT fine-tuning data extraction.
+<h3 align="center">
+  Governed Spec-Driven AI Delivery for Cursor and Claude Code
+</h3>
 
-100% free and open source. No paywalls.
+<p align="center">
+  <strong>Built on <a href="https://github.com/bmad-code-org/BMAD-METHOD">BMAD-METHOD</a> and <a href="https://github.com/github/spec-kit">Spec-Kit</a>.</strong><br>
+  <em>Runtime governance, mandatory audit loops, dashboard observability, and published npm installation in one flow.</em>
+</p>
+
+<p align="center">
+  <a href="LICENSE"><img src="https://img.shields.io/badge/License-MIT-yellow.svg" alt="License: MIT"></a>
+  <a href="https://nodejs.org"><img src="https://img.shields.io/badge/node-%3E%3D18.0.0-brightgreen" alt="Node.js Version"></a>
+</p>
 
 ---
 
-## Why BMAD-Speckit-SDD-Flow?
+## Why This Flow?
 
-Traditional AI tools do the thinking for you. BMAD-Speckit-SDD-Flow combines **BMAD Method** (agile, party-mode, multi-agent) with **Spec-Driven Development** (specify → plan → GAPS → tasks → TDD).
+Traditional AI tooling often stops at prompt orchestration. BMAD-Speckit-SDD-Flow turns that into a governed delivery pipeline: specify, plan, audit, enter implementation through readiness gates, run runtime governance during execution, then close out into scoring, dashboard, coach, and SFT data.
 
 <p align="center">
   <img src="docs/assets/readme-architecture-overview.final.svg" alt="BMAD-Speckit-SDD-Flow architecture overview" width="100%" />
 </p>
 
-What this adds in practice:
+### Key Capabilities
+
+- **5-layer delivery architecture**: Product Def → Epic Planning → Story Dev → Technical Implementation → Finish.
+- **Mandatory audit loops**: governed stages require review closure before continuing.
+- **Four-Signal implementation readiness**: implementation entry is blocked unless the readiness baseline covers `P0 Journey Coverage`, `Smoke E2E Readiness`, `Evidence Proof Chain`, and `Cross-Document Traceability`.
+- **Runtime gate loops and rerun gates**: governed routes can re-enter, rerun, or block with the same runtime truth path instead of silently bypassing failed close-out.
+- **Packet execution and closure evidence**: pass, required-fixes, blocked, and rerun outcomes are recorded as packet execution truth instead of host-local guesswork.
+- **Dashboard, coach, and SFT extraction**: runtime and scoring outputs feed the same downstream observability and optimization surfaces.
 
-- **Five-layer architecture** — Product Brief → PRD → Architecture → Epic/Story → speckit specify/plan/GAPS/tasks → TDD implement → PR + human review
-- **Mandatory audit loops** — Each stage requires code-review pass before proceeding
-- **Critical Auditor** — Dedicated challenger role, >60% share in party-mode
-- **Scoring system** — Multi-stage weighted scores, one-vote veto, AI Coach diagnosis
-- **SFT extraction** — Instruction-response pairs from low-score runs for fine-tuning
+> Image strategy: README assets live in `docs/assets/` and are tracked in Git. The package `README.md` is rendered on npm as GitHub Flavored Markdown, so keeping repository assets tracked and using repository-relative paths is the most reliable cross-surface strategy for GitHub and npm. Source: [About package README files](https://docs.npmjs.com/about-package-readme-files)
 
 ---
 
-## Quick Start
+## Runtime Governance At A Glance
 
-**Prerequisites**: [Node.js](https://nodejs.org) v18+
+- **Four-Signal readiness gate** runs before implementation entry and keeps readiness scoring separate from implementation scoring.
+- **Runtime gates loop** keeps workflow progression aligned with stage truth instead of letting hosts continue optimistically.
+- **Rerun gates** carry remediation and re-entry through the same governed path after fixes.
+- **Packet execution closure** preserves evidence for every governed execution outcome across Cursor and Claude Code hosts.
+- **Post-audit close-out** feeds scoring, dashboard, coach, and SFT extraction from the same runtime close-out path.
 
-```bash
-# Run the published root package without persisting it
-npx --yes --package bmad-speckit-sdd-flow bmad-speckit init . --ai cursor-agent --yes
+## Dashboard And MCP
 
-# Or create a new project
-npx --yes --package bmad-speckit-sdd-flow bmad-speckit init my-project --ai cursor-agent --yes
+- **Dashboard is default**: the published package supports runtime dashboard status, start/stop helpers, and runtime snapshot generation without any extra MCP setup.
+- **Runtime MCP is optional**: enable it only when you want runtime data exposed as agent tools via `--with-mcp`.
+- **Dashboard and runtime governance do not depend on MCP**: live dashboard, hooks, scoring projection, and runtime close-out all work without `.mcp.json`.
 
-# Verify installation
-npx --yes --package bmad-speckit-sdd-flow bmad-speckit check
-```
+Quick mental model:
 
-**Important boundary**:
+- `dashboard`: human-facing runtime and scoring visibility
+- `runtime-mcp`: explicit agent-tool surface over the same runtime data
 
-- The commands above execute the **single public root package** and expose the bundled `bmad-speckit` CLI
-- They are **not** the highest-confidence installation path for BMAD-Speckit-SDD-Flow custom consumer features on another machine
-- If your priority scenario is **installing into a consumer project on another machine without this repository checked out**, use the verified off-repo path in [Consumer Installation Guide](docs/how-to/consumer-installation.md)
-- In this repository, the verified off-repo path is: install the **root package release artifact** (`bmad-speckit-sdd-flow-<version>.tgz` or the equivalent published package), then run `npx bmad-speckit version` / `check` / `bmad-speckit-init`
+---
 
-### Install From A CI Artifact
+## Recommended npm Installation
 
-If you need to install this into a consumer project on another machine without cloning this repository, use the tarball produced by the CI `package` job:
+Ensure you have **[Node.js](https://nodejs.org) v18+** installed.
 
-1. Download the GitHub Actions artifact named `npm-packages-<commit-sha>`
-2. Extract the artifact and locate `bmad-speckit-sdd-flow-<version>.tgz`
-3. Install that root package tarball in the consumer project
-4. Run the verification commands before initializing host hooks
+### Recommended Off-Repo Install From npm
 
-```powershell
-cd D:\Dev\your-project
-npm install --save-dev .\bmad-speckit-sdd-flow-<version>.tgz
-npx bmad-speckit version
+Use the published root package directly. This is the current recommended path when you're installing into a consumer project without cloning this repository.
+
+This is the verified off-repo path for the published package contract.
+
+```bash
+npx --yes --package bmad-speckit-sdd-flow@latest bmad-speckit version
+npx --yes --package bmad-speckit-sdd-flow@latest bmad-speckit-init . --agent claude-code --full --no-package-json
+npx --yes --package bmad-speckit-sdd-flow@latest bmad-speckit-init . --agent cursor --full --no-package-json
+npx --yes --package bmad-speckit-sdd-flow@latest bmad-speckit check
+npx --yes --package bmad-speckit-sdd-flow@latest bmad-speckit dashboard-status
+```
+
+Why this is the recommended path:
+
+- it uses the published single public root package
+- it aligns both host surfaces explicitly
+- it preserves the non-invasive `--no-package-json` consumer install style
+- it matches the validated published package flow rather than an older bootstrap-only shortcut
+
+### Persistent Install In A Project
+
+If you want the package present in the consumer project's dependency tree:
+
+```bash
+npm install --save-dev bmad-speckit-sdd-flow@latest
+npx bmad-speckit-init . --agent claude-code --full --no-package-json
+npx bmad-speckit-init . --agent cursor --full --no-package-json
 npx bmad-speckit check
-npx bmad-speckit-init --agent claude-code
-npx bmad-speckit-init --agent cursor
 ```
 
-Use the root package tarball as the default install artifact. The CI artifact manifest is only there to help you verify filename, version, and checksum metadata.
+### Quick Bootstrap Path
+
+The faster bootstrap command still exists:
+
+```bash
+npx --yes --package bmad-speckit-sdd-flow@latest bmad-speckit init . --ai cursor-agent --yes
+```
+
+Treat that as a quick initializer, not the highest-confidence installation path for the full runtime governance surface. If you care about the latest published hooks, runtime governance, dashboard wiring, and dual-host alignment, use the recommended installation path above.
+
+> Need help choosing the next governed route? Run `/bmad-help` in your AI IDE. It evaluates `flow`, `contextMaturity`, `complexity`, and `implementationReadinessStatus` before recommending or blocking routes.
+
+### Alternative Deployments
 
-> **Not sure what to do?** Run `/bmad-help` in your AI IDE. 它会先识别你的 `flow / contextMaturity / complexity / implementationReadinessStatus`，再把路径标成 `recommended / allowed but not recommended / blocked`。See [Installation & Migration Guide](docs/how-to/migration.md) and [bmad-help 路由模型参考](docs/reference/bmad-help-routing-model.md) for details.
+<details>
+<summary><b>Install via CI Artifact (Consumer Projects)</b></summary>
+<br>
+If you are installing from a release artifact instead of npm registry:
 
-**One-line deploy**:
+1. Download the `npm-packages-<commit-sha>` artifact from GitHub Actions.
+2. Extract the `bmad-speckit-sdd-flow-<version>.tgz` tarball.
+3. Install and initialize:
+   ```bash
+   npx --yes --package ./bmad-speckit-sdd-flow-<version>.tgz bmad-speckit version
+   npx --yes --package ./bmad-speckit-sdd-flow-<version>.tgz bmad-speckit-init . --agent claude-code --full --no-package-json
+   npx --yes --package ./bmad-speckit-sdd-flow-<version>.tgz bmad-speckit-init . --agent cursor --full --no-package-json
+   ```
+   </details>
+
+<details>
+<summary><b>One-Line Deploy Scripts</b></summary>
+<br>
 
 ```powershell
 # Windows
@@ -86,93 +141,74 @@ pwsh scripts/setup.ps1 -Target <project-path>
 ```bash
 # WSL / Linux / macOS
 bash scripts/setup.sh -Target <project-path>
-# or: npm run setup:sh -- -Target <path>
 ```
 
-See [WSL / Shell scripts](docs/how-to/wsl-shell-scripts.md) for full shell script reference.
-
-面向消费项目的完整安装入口见 [Consumer Installation Guide](docs/how-to/consumer-installation.md)。如果你需要继续配置 provider 的 `baseUrl` / `apiKeyEnv` / `model`，也从这篇开始。
+</details>
 
-如果你后续需要安全卸载当前项目中的受管安装面，使用：
+<details>
+<summary><b>Safe Uninstallation</b></summary>
+<br>
+To remove managed installation surfaces in the current project:
 
 ```bash
-npx bmad-speckit uninstall
+npx --yes --package bmad-speckit-sdd-flow@latest bmad-speckit uninstall
 ```
 
-边界：
-
-- 只删除安装器受管条目
-- 不整删 `.cursor`、`.claude`、全局 skills 根目录
-- **禁止删除 `_bmad-output`**
+This only removes installer-managed entries. It does not delete `.cursor`, `.claude`, or global skills, and it never deletes `_bmad-output`.
 
-如果你后续确实需要把运行时信息通过工具接口暴露给 agent，再看 [Runtime MCP Installation](docs/how-to/runtime-mcp-installation.md)。该能力不是默认安装产物，需要显式启用。
+</details>
 
 ---
 
-## Built On
+## Architecture And Modules
 
-| Upstream                                                    | Purpose                                                      |
-| ----------------------------------------------------------- | ------------------------------------------------------------ |
-| [BMAD-METHOD](https://github.com/bmad-code-org/BMAD-METHOD) | Agile workflows, Party Mode, 34+ workflows                   |
-| [github/spec-kit](https://github.com/github/spec-kit)       | Spec-Driven Development (constitution, specify, plan, tasks) |
+### Core Components
 
-**Our extensions**: scoring, critical auditor, speckit-workflow audit loops, bmad-story-assistant, bmad-bug-assistant.
-
----
+| Component                   | Purpose                                                                                             |
+| :-------------------------- | :-------------------------------------------------------------------------------------------------- |
+| **`_bmad/`**                | Canonical source of truth for workflow modules, hooks, prompts, routing, and host-facing assets.    |
+| **`packages/scoring/`**     | Scoring engine, readiness drift evaluation, dashboard projection, coach inputs, and SFT extraction. |
+| **`dashboard`**             | Default runtime observability surface: live dashboard, runtime snapshots, and scoring projections.  |
+| **`runtime-mcp`**           | Optional MCP surface for agent tools over runtime data; enabled explicitly with `--with-mcp`.       |
+| **`speckit-workflow`**      | Specify -> Plan -> GAPS -> Tasks -> TDD with mandatory audit loops.                                 |
+| **`bmad-story-assistant`**  | Story lifecycle path: Create Story -> Party Mode -> Dev Story -> Implement.                         |
+| **`bmad-bug-assistant`**    | Bug lifecycle path: RCA -> Party Mode -> BUGFIX -> Implement.                                       |
+| **`bmad-standalone-tasks`** | Standalone execution path for TASKS or BUGFIX documents through governed subagent delivery.         |
 
-## Project Structure
+<details>
+<summary><b>View Folder Structure</b></summary>
 
-```
+```text
 BMAD-Speckit-SDD-Flow/
-├── _bmad/                              # BMAD core modules (single source of truth)
-│   ├── speckit/                        # Speckit module (commands, templates, workflows, scripts)
-│   ├── core/                           # Core BMAD functionality
-│   ├── bmm/, bmb/, cis/, tea/          # Other BMAD modules
-│   ├── scoring/                        # Scoring system
-│   └── _config/                        # Configuration files
-├── packages/                           # Source code (monorepo)
-│   ├── bmad-speckit/                   # Internal CLI workspace bundled into the root package
-│   └── scoring/                        # Scoring extensions (see [packages/scoring/README.md](packages/scoring/README.md))
-├── tests/                              # Test suite (acceptance/, epic-acceptance/)
-├── scripts/                            # Deployment/utility scripts only
-│   └── _config/                        # Project configuration (merged from former root config/)
-├── specs/                              # Story specifications (generated)
-├── docs/                               # Documentation (Diataxis)
-│   ├── tutorials/                      # Learning-oriented
-│   ├── how-to/                         # Task-oriented
-│   ├── explanation/                    # Understanding-oriented
-│   ├── reference/                      # Information-oriented
-│   └── sample/                         # Example documents
-└── package.json
+├── _bmad/                # Core modules and configuration
+├── packages/             # Monorepo packages (CLI, scoring)
+├── scripts/              # Setup and deployment utilities
+├── docs/                 # Diataxis-style documentation
+├── tests/                # Acceptance & epic testing
+└── specs/                # Generated story specs
 ```
 
-## Modules & Components
-
-| Component                 | Purpose                                                                                                                                                      |
-| ------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------ |
-| **\_bmad/**               | BMAD core (core, bmm, bmb, cis, tea, scoring)                                                                                                                |
-| **packages/scoring/**     | Scoring extensions: audit report parsing, score persistence, Coach diagnosis, Dashboard, SFT extraction                                                      |
-| **scripts/**              | Deployment/utility scripts: init-to-root, setup. Scoring 已整合进 bmad-speckit CLI，目标项目使用 `npx bmad-speckit score/coach/dashboard/sft-extract/scores` |
-| **speckit-workflow**      | specify → plan → GAPS → tasks → TDD with mandatory audits                                                                                                    |
-| **bmad-story-assistant**  | Create Story → Party-Mode → Dev Story → implement                                                                                                            |
-| **bmad-bug-assistant**    | Bug description → Party-Mode → BUGFIX doc                                                                                                                    |
-| **bmad-standalone-tasks** | Execute TASKS/BUGFIX docs via subagents                                                                                                                      |
+</details>
 
 ---
 
 ## Documentation
 
+Key entry points:
+
 - [Getting Started](docs/tutorials/getting-started.md)
 - [Consumer Installation Guide](docs/how-to/consumer-installation.md)
-- [Installation & Migration Guide](docs/how-to/migration.md)
+- [Runtime Dashboard Guide](docs/how-to/runtime-dashboard.md)
+- [Runtime MCP Installation](docs/how-to/runtime-mcp-installation.md)
+- [Provider Configuration](docs/how-to/provider-configuration.md)
 - [Cursor Setup](docs/how-to/cursor-setup.md)
 - [Claude Code Setup](docs/how-to/claude-code-setup.md)
-- [Provider Configuration](docs/how-to/provider-configuration.md)
-- [Runtime MCP Installation](docs/how-to/runtime-mcp-installation.md)
 - [WSL / Shell Scripts](docs/how-to/wsl-shell-scripts.md)
 
 ---
 
-## License
-
-MIT License — see [LICENSE](LICENSE) for details.
+<p align="center">
+  <a href="LICENSE">MIT License</a> •
+  <a href="https://github.com/bmad-code-org/BMAD-METHOD">BMAD-METHOD</a> •
+  <a href="https://github.com/github/spec-kit">Spec-Kit</a>
+</p>

package/README.zh-CN.md

@@ -0,0 +1,212 @@
+# BMAD-Speckit-SDD-Flow
+
+[English](README.md) | 简体中文
+
+<p align="center">
+  <img src="docs/assets/readme-slogan.final.svg" alt="BMAD-Speckit-SDD-Flow" width="100%" />
+</p>
+
+<h3 align="center">
+  面向 Cursor 与 Claude Code 的规范化 Spec-Driven AI 开发流程
+</h3>
+
+<p align="center">
+  <strong>基于 <a href="https://github.com/bmad-code-org/BMAD-METHOD">BMAD-METHOD</a> 与 <a href="https://github.com/github/spec-kit">Spec-Kit</a> 构建。</strong><br>
+  <em>把需求规范、审计流程、运行监控和评分反馈整合成一条完整的工程化交付链路。</em>
+</p>
+
+<p align="center">
+  <a href="LICENSE"><img src="https://img.shields.io/badge/License-MIT-yellow.svg" alt="License: MIT"></a>
+  <a href="https://nodejs.org"><img src="https://img.shields.io/badge/node-%3E%3D18.0.0-brightgreen" alt="Node.js Version"></a>
+</p>
+
+---
+
+## 这套流程要解决什么问题？
+
+很多 AI 工具只停留在提示词编排层面。BMAD-Speckit-SDD-Flow 把它推进成一条可管理的交付流水线：先写规范、再出计划、经过审计，确认 ready 后再进入实现阶段，执行过程中有运行时管控，最后统一输出评分、看板和训练数据。
+
+<p align="center">
+  <img src="docs/assets/readme-architecture-overview.final.svg" alt="架构总览" width="100%" />
+</p>
+
+### 主要特性
+
+- **五层交付架构**：Product Def → Epic Planning → Story Dev → Technical Implementation → Finish
+- **强制审计闭环**：每个治理阶段必须通过代码审查才能继续
+- **四信号就绪检查**：实现入口前需满足需求覆盖、冒烟测试就绪、证据链完整、文档可追溯
+- **运行时管控循环**：执行失败不会被静默跳过，而是沿同一条治理链路进入重试或阻塞
+- **执行证据链**：通过、需修复、阻塞、重跑等状态都会留下完整记录
+- **看板、诊断与训练数据**：运行结果自动汇总到可视化面板、问题诊断和模型微调数据集
+
+> **关于图片**：README 中的图片放在 `docs/assets/` 目录下并纳入 Git 管理。npm 包里的 README 会按 GitHub Flavored Markdown 渲染，因此"仓库内相对路径 + 已跟踪资源"是对 GitHub 和 npm 都最稳定的策略。来源：[About package README files](https://docs.npmjs.com/about-package-readme-files)
+
+---
+
+## 运行时管控一览
+
+- **四信号就绪检查**：在进入实现阶段前执行，与实现评分保持独立
+- **运行时管控循环**：工作流推进与阶段事实保持一致，不允许宿主"乐观继续"
+- **重跑检查点**：修复后沿同一条治理路径回到重试或重新进入
+- **执行闭环证据**：Cursor 与 Claude Code 两侧都保留完整的治理执行记录
+- **事后审计收口**：通过同一条运行时收口路径驱动评分、看板、诊断和训练数据提取
+
+## 看板与 MCP
+
+- **看板是默认能力**：发布包默认支持运行时看板状态查询、启停辅助、快照生成
+- **运行时 MCP 是可选能力**：只有在你希望把运行时数据暴露成 agent 工具接口时，才显式启用 `--with-mcp`
+- **看板和运行时管控不依赖 MCP**：实时看板、钩子、评分投影、运行时收口在没有 `.mcp.json` 的情况下也能工作
+
+简单理解：
+
+- `dashboard`：给人看的运行时/评分可视化
+- `runtime-mcp`：把同一份运行时数据暴露成 agent 工具接口
+
+---
+
+## 推荐的 npm 安装方式
+
+确保本机已安装 **[Node.js](https://nodejs.org) v18+**。
+
+### 推荐的 npm 离仓安装路径
+
+如果你是要把它装进一个消费项目，而不是修改本仓库源码，当前推荐直接使用已发布的根包：
+
+```bash
+npx --yes --package bmad-speckit-sdd-flow@latest bmad-speckit version
+npx --yes --package bmad-speckit-sdd-flow@latest bmad-speckit-init . --agent claude-code --full --no-package-json
+npx --yes --package bmad-speckit-sdd-flow@latest bmad-speckit-init . --agent cursor --full --no-package-json
+npx --yes --package bmad-speckit-sdd-flow@latest bmad-speckit check
+npx --yes --package bmad-speckit-sdd-flow@latest bmad-speckit dashboard-status
+```
+
+为什么推荐这条路径：
+
+- 它使用唯一公开发布的根包
+- 它显式对齐两侧宿主安装面
+- 它保留 `--no-package-json` 这种非侵入式消费安装风格
+- 它对应的是这次已经验证过的已发布 npm 路径，而不是旧的纯引导快捷入口
+
+### 持久安装到项目依赖树
+
+如果你希望把包写进消费项目的依赖树：
+
+```bash
+npm install --save-dev bmad-speckit-sdd-flow@latest
+npx bmad-speckit-init . --agent claude-code --full --no-package-json
+npx bmad-speckit-init . --agent cursor --full --no-package-json
+npx bmad-speckit check
+```
+
+### 快速引导路径
+
+更快的引导命令仍然保留：
+
+```bash
+npx --yes --package bmad-speckit-sdd-flow@latest bmad-speckit init . --ai cursor-agent --yes
+```
+
+但应把它理解成一个快速初始化入口，而不是完整运行时治理安装面的最高置信路径。如果你关心已发布钩子、运行时管控、看板接入和双宿主对齐，优先使用上面的推荐路径。
+
+> 不确定该走哪条治理路径时，在 AI IDE 中运行 `/bmad-help`。它会结合流程、上下文成熟度、复杂度和实现就绪状态做推荐或阻断。
+
+### 其他安装方式
+
+<details>
+<summary><b>通过 CI 产物安装到消费项目</b></summary>
+<br>
+如果你使用 release 产物，而不是直接从 npm registry 安装：
+
+1. 下载 GitHub Actions 产物 `npm-packages-<commit-sha>`
+2. 解压出 `bmad-speckit-sdd-flow-<version>.tgz`
+3. 执行：
+
+   ```bash
+   npx --yes --package ./bmad-speckit-sdd-flow-<version>.tgz bmad-speckit version
+   npx --yes --package ./bmad-speckit-sdd-flow-<version>.tgz bmad-speckit-init . --agent claude-code --full --no-package-json
+   npx --yes --package ./bmad-speckit-sdd-flow-<version>.tgz bmad-speckit-init . --agent cursor --full --no-package-json
+   ```
+
+</details>
+
+<details>
+<summary><b>一键部署脚本</b></summary>
+<br>
+
+```powershell
+# Windows
+pwsh scripts/setup.ps1 -Target <项目路径>
+```
+
+```bash
+# WSL / Linux / macOS
+bash scripts/setup.sh -Target <项目路径>
+```
+
+</details>
+
+<details>
+<summary><b>安全卸载</b></summary>
+<br>
+如果要移除当前项目里的受管安装面：
+
+```bash
+npx --yes --package bmad-speckit-sdd-flow@latest bmad-speckit uninstall
+```
+
+它只会删除安装器受管条目，不会整删 `.cursor`、`.claude` 或全局 skills，也不会删除 `_bmad-output`。
+
+</details>
+
+---
+
+## 架构与模块
+
+### 核心组件
+
+| 组件                        | 说明                                                                  |
+| :-------------------------- | :-------------------------------------------------------------------- |
+| **`_bmad/`**                | 工作流模块、钩子、提示词、路由与宿主侧资产的规范源                    |
+| **`packages/scoring/`**     | 评分引擎、就绪漂移评估、看板投影、诊断输入与训练数据提取              |
+| **`dashboard`**             | 默认运行时可观测层：实时看板、运行时快照、评分投影                    |
+| **`runtime-mcp`**           | 可选的 MCP 工具接口，通过 `--with-mcp` 显式启用                       |
+| **`speckit-workflow`**      | Specify → Plan → GAPS → Tasks → TDD，并带强制审计循环                 |
+| **`bmad-story-assistant`**  | Story 生命周期路径：Create Story → Party Mode → Dev Story → Implement |
+| **`bmad-bug-assistant`**    | Bug 生命周期路径：RCA → Party Mode → BUGFIX → Implement               |
+| **`bmad-standalone-tasks`** | 针对 TASKS 或 BUGFIX 文档的独立受管子代理执行路径                     |
+
+<details>
+<summary><b>查看目录结构</b></summary>
+
+```text
+BMAD-Speckit-SDD-Flow/
+├── _bmad/                # 核心模块与配置
+├── packages/             # Monorepo 包（CLI、评分）
+├── scripts/              # 安装与部署工具脚本
+├── docs/                 # Diataxis 风格文档
+├── tests/                # 验收测试与 epic 测试
+└── specs/                # 生成的 Story 规范
+```
+
+</details>
+
+---
+
+## 文档入口
+
+- [快速开始](docs/tutorials/getting-started.md)
+- [消费项目安装指南](docs/how-to/consumer-installation.md)
+- [运行时看板指南](docs/how-to/runtime-dashboard.md)
+- [运行时 MCP 安装](docs/how-to/runtime-mcp-installation.md)
+- [Provider 配置](docs/how-to/provider-configuration.md)
+- [Cursor 配置](docs/how-to/cursor-setup.md)
+- [Claude Code 配置](docs/how-to/claude-code-setup.md)
+- [WSL / Shell 脚本](docs/how-to/wsl-shell-scripts.md)
+
+---
+
+<p align="center">
+  <a href="LICENSE">MIT License</a> •
+  <a href="https://github.com/bmad-code-org/BMAD-METHOD">BMAD-METHOD</a> •
+  <a href="https://github.com/github/spec-kit">Spec-Kit</a>
+</p>