npm - jarvis-agent-factory - Versions diffs - 4.2.1 → 4.2.2 - Mend

jarvis-agent-factory 4.2.1 → 4.2.2

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Files changed (4) hide show

package/README.md +5 -4
package/dist/package.json +1 -1
package/dist/web/commands-reference.html +1125 -0
package/package.json +1 -1

package/README.md CHANGED Viewed

@@ -1,7 +1,7 @@
 # Jarvis Agent Factory · 贾维斯智能体工厂
 [![License: MIT](https://img.shields.io/badge/license-MIT-blue)](./LICENSE)
-[![Version](https://img.shields.io/badge/version-v4.2.0-green)](https://github.com/Wjl1224734792/Jarvis-Agent-Factory/releases)
+[![Version](https://img.shields.io/badge/version-v4.2.1-green)](https://github.com/Wjl1224734792/Jarvis-Agent-Factory/releases)
 [![npm](https://img.shields.io/npm/v/jarvis-agent-factory)](https://www.npmjs.com/package/jarvis-agent-factory)
 [![Visual Primitives MCP](https://img.shields.io/badge/DeepSeek-Visual%20Primitives%20MCP-purple)](https://github.com/Wjl1224734792/visual-primitives-mcp)
 <br>💡 **纯文本模型（如 DeepSeek）主力用户** → 搭配 [Visual Primitives MCP](https://github.com/Wjl1224734792/visual-primitives-mcp) 获得视觉理解能力
@@ -9,7 +9,7 @@
 AI 编程助手配置集 + MCP 编排引擎。从想法到交付的完整软件开发流水线，<br>**仅支持 Claude Code**。
-> **v4.2.0** — 孤立技能全接入(13个)+平台Agent全覆盖(GATE_AGENT_GUIDE+34)+评分权重框架(/ask K2)
+> **v4.2.1** — 生产就绪度加固：catch 硬化 + SSE 并发锁 + Gate C2 5次重试 + 15 MCP核心API集成测试 + 3 Expo Agent + 364全绿
 ## 快速开始
@@ -175,6 +175,7 @@ GITHUB_TOKEN=xxx       # GitHub 个人访问令牌（sync-github-releases 需要
 | 流水线看板 | `#/dashboard` | 会话列表（任务名/指令标签/Gate状态）· MCP 平台接入状态 · 置顶/归档/删除 |
 | 归档记录 | `#/archive` | 已归档运行记录 · 按任务名搜索过滤 · 恢复到看板 · 永久删除 |
 | 智能体配置 | `#/agents` | MCP 接入指示 · Agent 搜索/筛选 · 模型/思考等级配置 · 文件同步 |
+| 项目介绍 | `/commands-reference.html` | 项目概览 · 39条指令流程详解 · 71个Agent目录 · 流水线架构 · 使用推荐 |
 侧边栏实时显示 Claude Code 的 MCP 连接状态：绿点 = 已接入，灰点 = 未接入。
@@ -322,7 +323,7 @@ test-doc-writer → test-executor → remediation-expert
 | | Claude Code |
 |---|:--:|
-| Agents | 68 |
+| Agents | 71 |
 | Commands | 39 |
 | Skills | 34 |
 | Pipeline | 15 条流水线（full/lite/frontend/backend/refactor/hotfix/migrate/evaluate/debug/research/release/ask/simplify/trace/improve） |
@@ -366,7 +367,7 @@ test-doc-writer → test-executor → remediation-expert
 ## 命令流程图
-每个 Claude Code 命令的完整 Mermaid 流程图，展示 Gate 序列、Agent spawn 关系和并行/串行逻辑：
+每个 Claude Code 命令的完整介绍，包含项目概览、39条指令的详细流程（含 Gate 序列和 Agent spawn 关系）、71个智能体目录、流水线架构和使用推荐：[项目介绍页](https://github.com/Wjl1224734792/Jarvis-Agent-Factory/blob/main/web/public/commands-reference.html)
 | 分类 | 命令 | 流程图 | Gate 序列 |
 |------|------|--------|----------|

package/dist/package.json CHANGED Viewed

@@ -1,6 +1,6 @@
 {
   "name": "jarvis-agent-factory",
-  "version": "4.2.1",
+  "version": "4.2.2",
   "description": "Jarvis Agent Factory CLI — Claude Code 多智能体 AI 编程助手配置安装器 | Multi-agent AI coding assistant config installer for Claude Code",
   "keywords": [
     "jarvis",

package/dist/web/commands-reference.html ADDED Viewed

@@ -0,0 +1,1125 @@
+<!DOCTYPE html>
+<html lang="zh-CN">
+<head>
+<meta charset="UTF-8">
+<meta name="viewport" content="width=device-width, initial-scale=1.0">
+<title>Jarvis Agent Factory — 多智能体 AI 编程助手配置工厂</title>
+<style>
+  :root {
+    --bg: #0d1117; --card: #161b22; --border: #30363d; --text: #c9d1d9;
+    --muted: #8b949e; --accent: #58a6ff; --green: #3fb950; --orange: #d2991d;
+    --red: #f85149; --purple: #a371f7; --cyan: #39c5cf; --pink: #db61a2;
+  }
+  * { box-sizing:border-box; margin:0; padding:0 }
+  body { background:var(--bg); color:var(--text); font:14px/1.7 -apple-system,BlinkMacSystemFont,"Segoe UI",sans-serif; max-width:1200px; margin:0 auto; padding:20px }
+  /* Hero */
+  .hero { text-align:center; padding:48px 0 32px; border-bottom:1px solid var(--border); margin-bottom:32px }
+  .hero h1 { font-size:36px; color:#fff; margin-bottom:8px; letter-spacing:-0.5px }
+  .hero .tagline { font-size:18px; color:var(--muted); margin-bottom:20px }
+  .hero .version { display:inline-block; background:var(--accent); color:#fff; border-radius:4px; padding:2px 10px; font-size:13px; font-weight:600 }
+  /* Stats row */
+  .stats { display:flex; gap:16px; flex-wrap:wrap; justify-content:center; margin:24px 0 }
+  .stat { background:var(--card); border:1px solid var(--border); border-radius:10px; padding:16px 28px; text-align:center; min-width:100px; transition:border-color .2s }
+  .stat:hover { border-color:var(--accent) }
+  .stat .num { font-size:32px; font-weight:800; color:#fff; line-height:1.1 }
+  .stat .label { font-size:12px; color:var(--muted); margin-top:4px }
+  .stat.accent .num { color:var(--accent) }
+  .stat.green .num { color:var(--green) }
+  .stat.purple .num { color:var(--purple) }
+  .stat.orange .num { color:var(--orange) }
+  /* Nav */
+  .nav { position:sticky; top:0; background:rgba(13,17,23,0.92); backdrop-filter:blur(8px); padding:12px 0; border-bottom:1px solid var(--border); margin-bottom:28px; z-index:10; display:flex; gap:10px; flex-wrap:wrap; justify-content:center }
+  .nav a { color:var(--text); text-decoration:none; font-size:13px; padding:5px 10px; border-radius:4px; white-space:nowrap }
+  .nav a:hover { color:var(--accent); background:var(--card) }
+  /* Section */
+  h2 { font-size:22px; color:#fff; margin:40px 0 16px; padding-bottom:8px; border-bottom:1px solid var(--border); display:flex; align-items:center; gap:8px }
+  h2 .count { font-size:14px; color:var(--muted); font-weight:400 }
+  h3 { font-size:17px; color:var(--accent); margin:24px 0 10px }
+  h4 { font-size:15px; color:#fff; margin:16px 0 8px }
+  /* Intro block */
+  .intro { background:var(--card); border:1px solid var(--border); border-radius:10px; padding:20px 24px; margin:16px 0; line-height:1.8 }
+  .intro strong { color:#fff }
+  .intro code { background:rgba(88,166,255,0.12); color:var(--accent); padding:1px 5px; border-radius:3px; font-size:13px }
+  /* Concept grid */
+  .concept-grid { display:grid; grid-template-columns:repeat(auto-fill,minmax(260px,1fr)); gap:12px; margin:12px 0 }
+  .concept-card { background:var(--card); border:1px solid var(--border); border-radius:8px; padding:16px }
+  .concept-card .icon { font-size:24px; margin-bottom:6px }
+  .concept-card h4 { margin:0 0 6px; font-size:15px }
+  .concept-card p { font-size:13px; color:var(--muted); line-height:1.6 }
+  /* Command card */
+  .cmd-grid { display:grid; grid-template-columns:repeat(auto-fill,minmax(360px,1fr)); gap:12px; margin:12px 0 }
+  .cmd-card { background:var(--card); border:1px solid var(--border); border-radius:8px; padding:16px; transition:border-color .2s }
+  .cmd-card:hover { border-color:var(--accent) }
+  .cmd-card .cmd-name { font-size:16px; font-weight:700; color:var(--accent); font-family:monospace }
+  .cmd-card .cmd-desc { color:var(--text); margin:6px 0; font-size:13px; line-height:1.6 }
+  .cmd-card .cmd-flow { margin-top:8px; padding-top:8px; border-top:1px solid var(--border) }
+  .cmd-card .flow-step { display:flex; gap:6px; margin:3px 0; font-size:12px; align-items:baseline }
+  .cmd-card .flow-step .step-tag { background:var(--bg); border:1px solid var(--border); border-radius:3px; padding:0 5px; font-family:monospace; font-size:11px; white-space:nowrap; min-width:fit-content }
+  .cmd-card .flow-step .step-tag.green { color:var(--green); border-color:rgba(63,185,80,0.3) }
+  .cmd-card .flow-step .step-tag.orange { color:var(--orange); border-color:rgba(210,153,29,0.3) }
+  .cmd-card .flow-step .step-tag.accent { color:var(--accent); border-color:rgba(88,166,255,0.3) }
+  .cmd-card .flow-step .step-tag.red { color:var(--red); border-color:rgba(248,81,73,0.3) }
+  .cmd-card .flow-step .step-desc { color:var(--muted) }
+  .cmd-card .agents { margin-top:8px; font-size:11px; color:var(--muted) }
+  .cmd-card .agents span { color:var(--purple) }
+  /* Tags */
+  .tag { display:inline-block; border-radius:4px; padding:1px 7px; font-size:11px; font-weight:500 }
+  .tag-blue { background:rgba(88,166,255,0.15); color:var(--accent) }
+  .tag-green { background:rgba(63,185,80,0.15); color:var(--green) }
+  .tag-orange { background:rgba(210,153,29,0.15); color:var(--orange) }
+  .tag-purple { background:rgba(163,113,247,0.15); color:var(--purple) }
+  .tag-red { background:rgba(248,81,73,0.15); color:var(--red) }
+  .tag-cyan { background:rgba(57,197,207,0.15); color:var(--cyan) }
+  /* Agent table */
+  .agent-category { background:var(--card); border:1px solid var(--border); border-radius:8px; padding:16px; margin:12px 0 }
+  .agent-category h4 { margin:0 0 8px; font-size:15px; display:flex; align-items:center; gap:8px }
+  .agent-category h4 .count { font-size:12px; color:var(--muted); font-weight:400 }
+  .agent-list { display:flex; flex-wrap:wrap; gap:6px }
+  .agent-tag { background:var(--bg); border:1px solid var(--border); border-radius:4px; padding:2px 8px; font-size:12px; font-family:monospace; color:var(--text) }
+  /* Pipeline table */
+  .pipe-table { width:100%; border-collapse:collapse; margin:12px 0; font-size:13px }
+  .pipe-table th { text-align:left; color:var(--accent); padding:10px 12px; border-bottom:2px solid var(--border); font-weight:600 }
+  .pipe-table td { padding:8px 12px; border-bottom:1px solid var(--border) }
+  .pipe-table tr:hover { background:var(--card) }
+  .pipe-table .pipe-name { font-family:monospace; color:var(--green) }
+  /* Recommend */
+  .recommend { background:var(--card); border:1px solid var(--border); border-left:3px solid var(--accent); border-radius:0 8px 8px 0; padding:14px 18px; margin:10px 0; font-size:13px; line-height:1.7 }
+  .recommend.warn { border-left-color:var(--orange) }
+  .recommend.tip { border-left-color:var(--green) }
+  .recommend strong { color:#fff }
+  /* Scenario grid */
+  .scenario-grid { display:grid; grid-template-columns:repeat(auto-fill,minmax(340px,1fr)); gap:12px; margin:12px 0 }
+  .scenario-card { background:var(--card); border:1px solid var(--border); border-radius:8px; padding:14px 16px }
+  .scenario-card h4 { margin:0 0 6px; font-size:14px }
+  .scenario-card p { font-size:12px; color:var(--muted); line-height:1.6 }
+  .scenario-card code { background:rgba(88,166,255,0.12); color:var(--accent); padding:1px 5px; border-radius:3px; font-size:12px }
+  /* Footer */
+  .footer { text-align:center; color:var(--muted); font-size:12px; margin:40px 0 24px; padding-top:24px; border-top:1px solid var(--border) }
+  .footer a { color:var(--accent); text-decoration:none }
+  .footer a:hover { text-decoration:underline }
+  /* Responsive */
+  @media (max-width:768px) {
+    .cmd-grid,.concept-grid,.scenario-grid { grid-template-columns:1fr }
+    .hero h1 { font-size:26px }
+    .stats { gap:8px }
+    .stat { padding:12px 16px; min-width:70px }
+    .stat .num { font-size:24px }
+  }
+</style>
+</head>
+<body>
+<!-- ============================================================ -->
+<!-- HERO -->
+<!-- ============================================================ -->
+<div class="hero">
+  <h1>Jarvis Agent Factory</h1>
+  <div class="tagline">多智能体 AI 编程助手配置工厂 — 从想法到交付的完整软件开发流水线</div>
+  <span class="version">v4.2.1</span>
+</div>
+<div class="stats">
+  <div class="stat accent"><div class="num">39</div><div class="label">指令 Commands</div></div>
+  <div class="stat green"><div class="num">71</div><div class="label">智能体 Agents</div></div>
+  <div class="stat purple"><div class="num">34</div><div class="label">技能 Skills</div></div>
+  <div class="stat orange"><div class="num">15</div><div class="label">流水线 Pipelines</div></div>
+  <div class="stat"><div class="num">62</div><div class="label">闸门 Gates</div></div>
+</div>
+<!-- ============================================================ -->
+<!-- NAV -->
+<!-- ============================================================ -->
+<nav class="nav">
+  <a href="#overview">项目概览</a>
+  <a href="#concepts">核心概念</a>
+  <a href="#commands">指令体系</a>
+  <a href="#agents">智能体体系</a>
+  <a href="#pipelines">流水线架构</a>
+  <a href="#skills">技能体系</a>
+  <a href="#scenarios">使用推荐</a>
+  <a href="#quickstart">快速开始</a>
+</nav>
+<!-- ============================================================ -->
+<!-- 项目概览 -->
+<!-- ============================================================ -->
+<h2 id="overview">项目概览</h2>
+<div class="intro">
+  <p><strong>Jarvis Agent Factory</strong>（贾维斯智能体工厂）是一套面向 <strong>Claude Code</strong> 平台的多智能体配置工程。它将软件开发从"手动编码"提升为"编排驱动"——通过 39 条指令、71 个专业智能体、34 个技能和 15 条流水线，覆盖从需求澄清到发布上线的完整软件生命周期。</p>
+  <br>
+  <p>核心思路：<strong>FSM 硬约束闸门系统</strong>（Gate A→E）强制每个阶段完成必要产出后才能推进，确保工程严谨性；同时 <code>/auto</code> 智能路由自动检测任务类型并跳过无关闸门，兼顾日常效率。</p>
+  <br>
+  <p>安装方式：<code>npm i -g jarvis-agent-factory</code> → <code>jarvis init -y</code> → Claude Code 重启后引擎自动拉起。零原生依赖（Node 22.5+ 内置 node:sqlite），安装秒级完成。</p>
+</div>
+<!-- ============================================================ -->
+<!-- 核心概念 -->
+<!-- ============================================================ -->
+<h2 id="concepts">核心概念</h2>
+<div class="concept-grid">
+  <div class="concept-card">
+    <div class="icon">&#x1F4CB;</div>
+    <h4>指令 Command</h4>
+    <p>用户交互入口，以 <code>/</code> 开头（如 <code>/jarvis</code>、<code>/auto</code>）。每条指令对应一条流水线，定义从需求到交付的完整流程。共 39 条。</p>
+  </div>
+  <div class="concept-card">
+    <div class="icon">&#x1F916;</div>
+    <h4>智能体 Agent</h4>
+    <p>专业化 AI 角色，在流水线各阶段由编排者 spawn 执行。按职能分 6 类：实现(25)、测试(15)、审查(16)、规划(4)、架构(4)、支撑(7)。共 71 个。</p>
+  </div>
+  <div class="concept-card">
+    <div class="icon">&#x1F4AA;</div>
+    <h4>技能 Skill</h4>
+    <p>可复用的行为指南和能力模块，由 Agent 在任务执行中加载。涵盖需求、规划、实现、质量、测试、安全、文档等 12 个领域。共 34 个。</p>
+  </div>
+  <div class="concept-card">
+    <div class="icon">&#x1F3ED;</div>
+    <h4>流水线 Pipeline</h4>
+    <p>预定义的 Gate 序列，每种开发场景对应一条。full（13门完整）和 lite（自适应跳过）是两条核心流水线，另有 13 条专业流水线。共 15 条。</p>
+  </div>
+  <div class="concept-card">
+    <div class="icon">&#x1F6AA;</div>
+    <h4>闸门 Gate</h4>
+    <p>FSM 硬约束检查点，当前 Gate 通过后才能推进到下一 Gate。引擎通过 <code>gate_check</code> 和 <code>gate_enforce</code> 强制执行操作权限表，禁止跳过或回退。</p>
+  </div>
+  <div class="concept-card">
+    <div class="icon">&#x2699;</div>
+    <h4>引擎 Engine</h4>
+    <p>MCP stdio 自动拉起的编排引擎（端口 3456），内置 SQLite WAL 模式、SSE 实时推送、FSM Gate 系统。负责会话管理、Gate 执行、Agent 调度和文档产物追踪。</p>
+  </div>
+</div>
+<!-- ============================================================ -->
+<!-- 指令体系 -->
+<!-- ============================================================ -->
+<h2 id="commands">指令体系 <span class="count">— 39 条指令，8 个分类</span></h2>
+<div class="intro">
+  <p>每条指令是一条完整的开发流水线入口。输入 <code>/指令名 任务描述</code> 即可启动。引擎自动推进 Gate 序列，在对应阶段 spawn 专业 Agent 执行任务。<strong>不确定用什么指令？直接用 <code>/auto</code>。</strong></p>
+</div>
+<!-- 编排入口 -->
+<h3>一、编排入口（2条）</h3>
+<div class="recommend">
+  <strong>日常默认入口：</strong><code style="color:var(--accent)">/auto</code> — 自动检测 12 种任务类型 → 路由最优流水线 → 智能跳过无关 Gate → 按复杂度分配 Team/Subagent。<strong>99% 的情况用它就够了。</strong>
+</div>
+<div class="cmd-grid">
+<div class="cmd-card">
+  <div class="cmd-name">/jarvis</div>
+  <div class="cmd-desc"><strong>全流程严格模式</strong> — 13 Gate 全部强制执行，适合中大型功能开发。从需求到发布的完整生命周期。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Gate A</span><span class="step-desc">需求澄清：分析用户输入，产出 REQ-XXX 需求文档</span></div>
+    <div class="flow-step"><span class="step-tag green">Gate B-DDD</span><span class="step-desc">领域驱动分析：识别聚合/实体/值对象/领域服务</span></div>
+    <div class="flow-step"><span class="step-tag green">Gate B-BDD</span><span class="step-desc">行为驱动场景：Gherkin Given/When/Then 验收场景</span></div>
+    <div class="flow-step"><span class="step-tag green">Gate B-TDD</span><span class="step-desc">测试驱动分解：TASK-XXX 映射 REQ-XXX，Red→Green→Refactor</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate B1</span><span class="step-desc">架构评审：spawn 对应架构师，产出架构评审报告</span></div>
+    <div class="flow-step"><span class="step-tag accent">Gate C</span><span class="step-desc">执行规划：planner 产出 parallel_batches 并行任务分组</span></div>
+    <div class="flow-step"><span class="step-tag accent">Gate C-impl</span><span class="step-desc">并行实现：Team 模式 spawn 多个实现 Agent 并行编码</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate C1</span><span class="step-desc">质量门：Lint + Type-check + Build + Deps Audit 全部通过</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate C1.5</span><span class="step-desc">视觉验证：截图对比，确保 UI 与设计一致（条件性）</span></div>
+    <div class="flow-step"><span class="step-tag red">Gate C2</span><span class="step-desc">测试验证：Team 模式 spawn 测试 Agent，最多 5 次重试</span></div>
+    <div class="flow-step"><span class="step-tag red">Gate D</span><span class="step-desc">评审：Team 模式 spawn 审查 Agent，产出分级审查报告</span></div>
+    <div class="flow-step"><span class="step-tag accent">Gate E</span><span class="step-desc">发布上线：质量重检 → 版本递增 → CHANGELOG → commit+tag+push</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>task-design / planner / frontend-architect / backend-architect / 实现类 Agent / 测试类 Agent / 审查类 Agent</div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/auto</div>
+  <div class="cmd-desc"><strong>智能路由模式</strong> — 自动检测 12 种任务类型（新功能/Bug/重构/热修复/审查/调研/调试/简化/改进/迁移/发布/小修改），路由最优流水线，智能跳过无关 Gate。按复杂度（轻/中/重）分配 Team 或 Subagent。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag accent">Step 1</span><span class="step-desc">任务类型检测：分析输入，匹配 12 种已知类型</span></div>
+    <div class="flow-step"><span class="step-tag accent">Step 2</span><span class="step-desc">流水线路由：选择 full / refactor / hotfix / research / debug / simplify / improve / migrate / release 流水线</span></div>
+    <div class="flow-step"><span class="step-tag green">Step 3</span><span class="step-desc">Gate 跳过：检测无关 Gate（如纯后端跳过 C1.5），调用 gate_jump</span></div>
+    <div class="flow-step"><span class="step-tag orange">Step 4</span><span class="step-desc">复杂度评估：轻→Subagent / 中→Team 2-3 / 重→Team 4-6</span></div>
+    <div class="flow-step"><span class="step-tag red">Step 5</span><span class="step-desc">执行 + 质量门 + 测试 + 评审：按流水线 Gate 推进</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>按路由结果动态选择</div>
+</div>
+</div>
+<!-- 平台开发 -->
+<h3>二、平台开发（8条）</h3>
+<div class="recommend tip">
+  <strong>推荐：</strong>移动端/跨端开发使用对应平台命令，它们使用平台专属 Agent（如 android-dev-expert、flutter-ui-expert）并强制执行 C1.5 视觉验证 Gate。
+</div>
+<div class="cmd-grid">
+<div class="cmd-card">
+  <div class="cmd-name">/frontend</div>
+  <div class="cmd-desc">前端开发生命周期 — C1.5 视觉验证强制。覆盖 React/Vue/Angular 等前端框架的完整开发流程。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Gate A</span><span class="step-desc">需求澄清 → 前端需求文档</span></div>
+    <div class="flow-step"><span class="step-tag green">Gate B</span><span class="step-desc">DDD→BDD→TDD 任务分解</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate B1</span><span class="step-desc">frontend-architect 架构评审</span></div>
+    <div class="flow-step"><span class="step-tag accent">Gate C→C-impl</span><span class="step-desc">planner 规划 → frontend-dev/ui/state-expert 并行实现</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate C1→C1.5</span><span class="step-desc">质量门 → <strong>视觉验证强制</strong>（截图对比）</span></div>
+    <div class="flow-step"><span class="step-tag red">Gate C2→D→E</span><span class="step-desc">测试 → 评审 → 发布</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>frontend-architect / frontend-dev-expert / frontend-ui-expert / frontend-state-expert / frontend-test-expert / frontend-review-expert</div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/backend</div>
+  <div class="cmd-desc">后端开发生命周期 — 跳过 C1.5 视觉验证（纯后端无 UI）。覆盖 API 设计、数据库架构、业务逻辑的完整开发流程。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Gate A</span><span class="step-desc">需求澄清 → 后端需求文档</span></div>
+    <div class="flow-step"><span class="step-tag green">Gate B</span><span class="step-desc">DDD→BDD→TDD 任务分解</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate B1</span><span class="step-desc">backend-architect / database-architect 架构评审</span></div>
+    <div class="flow-step"><span class="step-tag accent">Gate C→C-impl</span><span class="step-desc">planner 规划 → backend-dev/api/logic/data-expert 并行实现</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate C1</span><span class="step-desc">质量门（跳过 C1.5 视觉验证）</span></div>
+    <div class="flow-step"><span class="step-tag red">Gate C2→D→E</span><span class="step-desc">测试 → 评审 → 发布</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>backend-architect / database-architect / backend-dev-expert / backend-api-expert / backend-logic-expert / backend-data-expert / backend-test-expert / backend-review-expert</div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/android</div>
+  <div class="cmd-desc">Android 原生开发生命周期 — Kotlin/Jetpack Compose。C1.5 视觉验证强制。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Gate A→B</span><span class="step-desc">需求澄清 → 任务分解</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate B1</span><span class="step-desc">android-dev-expert 架构评审</span></div>
+    <div class="flow-step"><span class="step-tag accent">Gate C→C-impl</span><span class="step-desc">planner → android-dev/ui/state-expert 并行实现</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate C1→C1.5</span><span class="step-desc">质量门 → 视觉验证（C1.5强制）</span></div>
+    <div class="flow-step"><span class="step-tag red">Gate C2→D→E</span><span class="step-desc">测试（android-test-expert）→ 评审 → 发布</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>android-dev-expert / android-ui-expert / android-state-expert / android-test-expert / android-review-expert</div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/ios</div>
+  <div class="cmd-desc">iOS 原生开发生命周期 — Swift/SwiftUI。C1.5 视觉验证强制。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Gate A→B</span><span class="step-desc">需求澄清 → 任务分解</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate B1</span><span class="step-desc">ios-dev-expert 架构评审</span></div>
+    <div class="flow-step"><span class="step-tag accent">Gate C→C-impl</span><span class="step-desc">planner → ios-dev/ui/state-expert 并行实现</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate C1→C1.5</span><span class="step-desc">质量门 → 视觉验证（C1.5强制）</span></div>
+    <div class="flow-step"><span class="step-tag red">Gate C2→D→E</span><span class="step-desc">测试（ios-test-expert）→ 评审 → 发布</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>ios-dev-expert / ios-ui-expert / ios-state-expert / ios-test-expert / ios-review-expert</div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/flutter</div>
+  <div class="cmd-desc">Flutter 跨端开发生命周期 — Dart。一套代码 Android+iOS+Web。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Gate A→B</span><span class="step-desc">需求澄清 → 任务分解</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate B1</span><span class="step-desc">flutter-dev-expert 架构评审</span></div>
+    <div class="flow-step"><span class="step-tag accent">Gate C→C-impl</span><span class="step-desc">planner → flutter-dev/ui/state-expert 并行实现</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate C1→C1.5</span><span class="step-desc">质量门 → 视觉验证</span></div>
+    <div class="flow-step"><span class="step-tag red">Gate C2→D→E</span><span class="step-desc">测试（flutter-test-expert）→ 评审 → 发布</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>flutter-dev-expert / flutter-ui-expert / flutter-state-expert / flutter-test-expert / flutter-review-expert</div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/expo</div>
+  <div class="cmd-desc">Expo 跨端开发生命周期 — React Native + Expo SDK。快速原型和跨端发布。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Gate A→B</span><span class="step-desc">需求澄清 → 任务分解</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate B1</span><span class="step-desc">expo-dev-expert 架构评审</span></div>
+    <div class="flow-step"><span class="step-tag accent">Gate C→C-impl</span><span class="step-desc">planner → expo-dev/ui/state-expert 并行实现</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate C1→C1.5</span><span class="step-desc">质量门 → 视觉验证</span></div>
+    <div class="flow-step"><span class="step-tag red">Gate C2→D→E</span><span class="step-desc">测试（expo-test-expert）→ 评审 → 发布</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>expo-dev-expert / expo-ui-expert / expo-state-expert / expo-test-expert / expo-review-expert</div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/react-native</div>
+  <div class="cmd-desc">React Native 跨端开发生命周期 — JavaScript/TypeScript。Android+iOS 原生桥接。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Gate A→B</span><span class="step-desc">需求澄清 → 任务分解</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate B1</span><span class="step-desc">react-native-dev-expert 架构评审</span></div>
+    <div class="flow-step"><span class="step-tag accent">Gate C→C-impl</span><span class="step-desc">planner → react-native-dev/ui/state-expert 并行实现</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate C1→C1.5</span><span class="step-desc">质量门 → 视觉验证</span></div>
+    <div class="flow-step"><span class="step-tag red">Gate C2→D→E</span><span class="step-desc">测试 → 评审 → 发布</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>react-native-dev-expert / react-native-ui-expert / react-native-state-expert / react-native-test-expert / react-native-review-expert</div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/taro</div>
+  <div class="cmd-desc">Taro 小程序/H5 跨端开发生命周期 — 一套代码多端适配（微信/支付宝/百度/字节小程序 + H5）。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Gate A→B</span><span class="step-desc">需求澄清 → 任务分解</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate B1</span><span class="step-desc">taro-dev-expert 架构评审</span></div>
+    <div class="flow-step"><span class="step-tag accent">Gate C→C-impl</span><span class="step-desc">planner → taro-dev/ui/state-expert 并行实现</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate C1→C1.5</span><span class="step-desc">质量门 → 视觉验证（多端适配检查）</span></div>
+    <div class="flow-step"><span class="step-tag red">Gate C2→D→E</span><span class="step-desc">测试（taro-test-expert）→ 评审 → 发布</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>taro-dev-expert / taro-ui-expert / taro-state-expert / taro-test-expert / taro-review-expert</div>
+</div>
+</div>
+<!-- 需求与设计 -->
+<h3>三、需求与设计（5条）</h3>
+<div class="cmd-grid">
+<div class="cmd-card">
+  <div class="cmd-name">/ask</div>
+  <div class="cmd-desc"><strong>需求探询</strong> — 4 模式自适应：Interview（模糊→Socratic 追问澄清）、Direct（清晰→直接分析）、Consensus（ralplan 式 3 阶段共识：Planner→Architect→Critic）、Review（Critic 评估现有计划）。含 5 维度评分矩阵（BV/EF/RS/UI/DC）。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Gate K0</span><span class="step-desc">模式选择：分析输入清晰度，选择 Interview/Direct/Consensus/Review</span></div>
+    <div class="flow-step"><span class="step-tag green">Gate K1</span><span class="step-desc">信息收集：需求背景、业务价值、功能范围、用户场景</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate K2</span><span class="step-desc">多维度分析：BV(业务价值) / EF(工程可行) / RS(风险安全) / UI(用户体验) / DC(交付信心) 评分</span></div>
+    <div class="flow-step"><span class="step-tag accent">Gate K3</span><span class="step-desc">产出：需求分析报告 + 流水线路由建议</span></div>
+  </div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/research</div>
+  <div class="cmd-desc"><strong>深度研究</strong> — 5 阶段技术调研：课题定义→信息收集→多维分析→假设验证→研究报告。竞争假设+证据矩阵，适合技术选型和方案研究。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Gate RS0</span><span class="step-desc">课题定义：明确研究范围、目标和成功标准</span></div>
+    <div class="flow-step"><span class="step-tag green">Gate RS1</span><span class="step-desc">信息收集：多源数据采集（文档/社区/案例/基准测试）</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate RS2</span><span class="step-desc">多维分析：技术成熟度/生态/性能/安全性/维护成本</span></div>
+    <div class="flow-step"><span class="step-tag red">Gate RS3</span><span class="step-desc">假设验证：证据矩阵 + 竞争假设评估</span></div>
+    <div class="flow-step"><span class="step-tag accent">Gate RS4</span><span class="step-desc">研究报告：结论 + 推荐方案 + 实施路线图</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>docs-research-expert / external-resource-expert</div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/task-ddd</div>
+  <div class="cmd-desc"><strong>DDD 领域驱动分析</strong> — 产出聚合/实体/值对象/领域服务/聚合行为清单 + 流水线路由建议。对应 Gate B-DDD。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Step 1</span><span class="step-desc">业务领域识别：界定核心域/支撑域/通用域</span></div>
+    <div class="flow-step"><span class="step-tag green">Step 2</span><span class="step-desc">聚合设计：识别聚合根/实体/值对象，定义不变性约束</span></div>
+    <div class="flow-step"><span class="step-tag orange">Step 3</span><span class="step-desc">领域服务：跨聚合业务流程建模</span></div>
+    <div class="flow-step"><span class="step-tag accent">Step 4</span><span class="step-desc">产出：DDD 分析文档 + 聚合行为清单 + 路由建议</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>task-design</div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/task-bdd</div>
+  <div class="cmd-desc"><strong>BDD 行为驱动场景</strong> — Gherkin Given/When/Then 格式，编写验收场景。对应 Gate B-BDD。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Step 1</span><span class="step-desc">功能提取：从需求文档提取用户故事</span></div>
+    <div class="flow-step"><span class="step-tag green">Step 2</span><span class="step-desc">场景编写：Given(前置)→When(操作)→Then(预期结果)</span></div>
+    <div class="flow-step"><span class="step-tag orange">Step 3</span><span class="step-desc">边界覆盖：正常流程+异常路径+边界条件</span></div>
+    <div class="flow-step"><span class="step-tag accent">Step 4</span><span class="step-desc">产出：BDD 场景文档（可执行验收测试用例）</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>task-design</div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/task-tdd</div>
+  <div class="cmd-desc"><strong>TDD 测试驱动任务分解</strong> — Red→Green→Refactor 循环，TASK-XXX 映射 REQ-XXX。对应 Gate B-TDD。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Step 1</span><span class="step-desc">任务拆分：按 TDD 循环（Red 先写失败测试→Green 最小实现→Refactor 重构）</span></div>
+    <div class="flow-step"><span class="step-tag green">Step 2</span><span class="step-desc">映射建立：TASK-XXX → REQ-XXX 可追溯映射</span></div>
+    <div class="flow-step"><span class="step-tag orange">Step 3</span><span class="step-desc">测试骨架：为每个 TASK 生成测试框架代码</span></div>
+    <div class="flow-step"><span class="step-tag accent">Step 4</span><span class="step-desc">产出：TDD 任务包（测试骨架 + 实现任务 + 重构清单）</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>task-design</div>
+</div>
+</div>
+<!-- 工程操作 -->
+<h3>四、工程操作（8条）</h3>
+<div class="cmd-grid">
+<div class="cmd-card">
+  <div class="cmd-name">/refactor</div>
+  <div class="cmd-desc"><strong>代码重构安全网</strong> — 5 Gate 保护：定义边界→基线测试+覆盖率→执行重构→行为漂移检测→报告。失败自动回滚。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Gate R1</span><span class="step-desc">定义边界：明确重构范围、不变性约束、成功标准</span></div>
+    <div class="flow-step"><span class="step-tag green">Gate R2</span><span class="step-desc">基线建立：运行现有测试+记录覆盖率基线</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate R3</span><span class="step-desc">执行重构：逐文件重构，每步验证测试通过</span></div>
+    <div class="flow-step"><span class="step-tag red">Gate R4</span><span class="step-desc">行为漂移检测：对比重构前后测试结果和覆盖率</span></div>
+    <div class="flow-step"><span class="step-tag accent">Gate R5</span><span class="step-desc">报告：变更清单 + 风险点 + 回归测试结果</span></div>
+  </div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/hotfix</div>
+  <div class="cmd-desc"><strong>紧急热修复</strong> — 4 Gate 紧急协议：声明+审批→最小化修复→快速验证+回滚预案→事后审计。需要审批确认。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag red">Gate H0</span><span class="step-desc">紧急声明：评估影响范围+紧急程度+审批确认</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate H1</span><span class="step-desc">最小化修复：只修复根因，不做额外改动</span></div>
+    <div class="flow-step"><span class="step-tag green">Gate H2</span><span class="step-desc">快速验证：核心流程回归 + 回滚预案确认</span></div>
+    <div class="flow-step"><span class="step-tag accent">Gate H3</span><span class="step-desc">事后审计：根因分析 + 预防措施 + 审计报告</span></div>
+  </div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/migrate</div>
+  <div class="cmd-desc"><strong>框架迁移</strong> — 4 Gate：定义验证规则→应用迁移→编译验证→自动修复 Lint（最多 2 轮）。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Gate M1</span><span class="step-desc">迁移规则：定义 API 变更映射 + 验证标准</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate M2</span><span class="step-desc">应用迁移：批量替换 API 调用 + 更新依赖</span></div>
+    <div class="flow-step"><span class="step-tag accent">Gate M3</span><span class="step-desc">编译验证：确保项目编译通过</span></div>
+    <div class="flow-step"><span class="step-tag red">Gate M4</span><span class="step-desc">自动修复：运行 Lint --fix，最多 2 轮</span></div>
+  </div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/debug</div>
+  <div class="cmd-desc"><strong>调试诊断</strong> — 5 Gate 交互式诊断：收集信息→最小复现→调试会话→交互诊断→报告+修复方案。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Gate D0</span><span class="step-desc">信息收集：错误日志/堆栈/环境信息/复现步骤</span></div>
+    <div class="flow-step"><span class="step-tag green">Gate D1</span><span class="step-desc">最小复现：隔离问题，构造最小可复现用例</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate D2</span><span class="step-desc">调试会话：断点/日志/变量追踪，定位根因</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate D3</span><span class="step-desc">交互诊断：多轮验证假设，排除干扰因素</span></div>
+    <div class="flow-step"><span class="step-tag accent">Gate D4</span><span class="step-desc">报告：根因 + 修复方案 + 预防建议</span></div>
+  </div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/trace</div>
+  <div class="cmd-desc"><strong>贝叶斯因果追踪</strong> — 5 Gate：问题框架→竞争假设(2-5个)→证据收集→贝叶斯更新(>70%置信度)→解决方案。适合复杂根因分析。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Gate T0</span><span class="step-desc">问题框架：定义现象、时间线、已知事实</span></div>
+    <div class="flow-step"><span class="step-tag green">Gate T1</span><span class="step-desc">假设生成：2-5 个竞争假设，每个标注先验概率</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate T2</span><span class="step-desc">证据收集：为每个假设收集支持/反对证据</span></div>
+    <div class="flow-step"><span class="step-tag red">Gate T3</span><span class="step-desc">贝叶斯更新：用证据更新假设概率，置信度>70% 接受方案</span></div>
+    <div class="flow-step"><span class="step-tag accent">Gate T4</span><span class="step-desc">解决方案：基于最可能假设的修复方案</span></div>
+  </div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/evaluate</div>
+  <div class="cmd-desc"><strong>技术评估</strong> — 4 Gate：定义标准+权重→生成原型→收集指标→多维度报告+推荐方案。适合技术选型。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Gate E0</span><span class="step-desc">评估标准：定义维度+权重（性能/生态/学习成本/维护性）</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate E1</span><span class="step-desc">原型生成：为每个候选方案生成概念验证原型</span></div>
+    <div class="flow-step"><span class="step-tag accent">Gate E2</span><span class="step-desc">指标收集：运行基准测试，收集量化数据</span></div>
+    <div class="flow-step"><span class="step-tag green">Gate E3</span><span class="step-desc">报告：多维雷达图 + 加权评分 + 推荐方案</span></div>
+  </div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/simplify</div>
+  <div class="cmd-desc"><strong>代码简化</strong> — 4 Gate 五桶分析：死代码/重复/过度抽象/AI痕迹/复杂性。并行简化→回归验证→报告。失败自动回滚。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Gate S0</span><span class="step-desc">五桶分析：死代码→重复代码→过度抽象→AI痕迹→高复杂度，分类标记</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate S1</span><span class="step-desc">并行简化：5 个维度并行处理，每步安全检查和回归验证</span></div>
+    <div class="flow-step"><span class="step-tag red">Gate S2</span><span class="step-desc">回归验证：全部测试通过+覆盖率不降，失败自动回滚</span></div>
+    <div class="flow-step"><span class="step-tag accent">Gate S3</span><span class="step-desc">报告：简化前后对比（行数/圈复杂度/重复率）</span></div>
+  </div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/improve</div>
+  <div class="cmd-desc"><strong>度量驱动迭代改进</strong> — 5 Gate + 迭代循环：量化目标→研究→计划→执行+benchmark→评估→迭代/停止。支持 IM4→IM1 循环直到达标。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Gate IM0</span><span class="step-desc">量化目标：定义目标指标+baseline测量+成功阈值</span></div>
+    <div class="flow-step"><span class="step-tag green">Gate IM1</span><span class="step-desc">研究：分析瓶颈，生成优化假设</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate IM2</span><span class="step-desc">计划：选择最佳假设，制定实施计划</span></div>
+    <div class="flow-step"><span class="step-tag accent">Gate IM3</span><span class="step-desc">执行：实施优化 + 运行 benchmark 测量</span></div>
+    <div class="flow-step"><span class="step-tag red">Gate IM4</span><span class="step-desc">评估：达标→停止；未达标→回 Gate IM1 继续迭代</span></div>
+  </div>
+</div>
+</div>
+<!-- 审查与修复 -->
+<h3>五、审查与修复（3条）</h3>
+<div class="cmd-grid">
+<div class="cmd-card">
+  <div class="cmd-name">/review</div>
+  <div class="cmd-desc"><strong>只读代码审查</strong> — 不修改文件。spawn 多个审查 Agent 并行收集 findings（每条含文件/行号/依据/严重度），产出审查报告+严重度分级。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Step 1</span><span class="step-desc">范围确定：选定审查文件和关注维度</span></div>
+    <div class="flow-step"><span class="step-tag orange">Step 2</span><span class="step-desc">并行审查：spawn 审查 Agent 并行扫描（安全/性能/风格/架构）</span></div>
+    <div class="flow-step"><span class="step-tag red">Step 3</span><span class="step-desc">发现汇总：合并 findings + 去重 + 严重度分级（P0/P1/P2/P3）</span></div>
+    <div class="flow-step"><span class="step-tag accent">Step 4</span><span class="step-desc">报告输出：审查报告（只读，不修改代码）</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>frontend-review-expert / backend-review-expert / security-review-expert / perf-review-expert / project-review-expert / diff-review-expert / change-review-expert / qa-review-expert</div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/review-fix</div>
+  <div class="cmd-desc"><strong>审查修复优化闭环</strong> — ①初审(只读)→②修复规划→③执行修复→④验证(质量重检)→⑤复审。最多 2 轮闭环。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Step 1</span><span class="step-desc">初审：spawn 审查 Agent 并行审查（同 /review）</span></div>
+    <div class="flow-step"><span class="step-tag orange">Step 2</span><span class="step-desc">修复规划：remediation-planner 制定修复方案，按严重度排序</span></div>
+    <div class="flow-step"><span class="step-tag accent">Step 3</span><span class="step-desc">执行修复：spawn 对应实现 Agent 逐项修复</span></div>
+    <div class="flow-step"><span class="step-tag red">Step 4</span><span class="step-desc">质量重检：Lint+Type-check+Build+测试 全部通过</span></div>
+    <div class="flow-step"><span class="step-tag orange">Step 5</span><span class="step-desc">复审：重新审查，≤2轮。仍失败→ABORT</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>review-fix-optimize / remediation-planner / review-only / 实现类+审查类 Agent</div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/bug-fix</div>
+  <div class="cmd-desc"><strong>Bug 修复闭环</strong> — 浏览器复现→根因定位→修复→浏览器验证。涉及前端/交互类 Bug 的 7 步闭环，集成 Playwright MCP。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Step 1</span><span class="step-desc">Bug 描述：收集复现步骤+期望行为+实际行为</span></div>
+    <div class="flow-step"><span class="step-tag orange">Step 2</span><span class="step-desc">浏览器复现：Playwright 打开页面，按步骤操作确认 Bug</span></div>
+    <div class="flow-step"><span class="step-tag orange">Step 3</span><span class="step-desc">根因定位：代码追踪→假设验证→确定根因</span></div>
+    <div class="flow-step"><span class="step-tag accent">Step 4</span><span class="step-desc">修复实现：最小化代码修改</span></div>
+    <div class="flow-step"><span class="step-tag green">Step 5</span><span class="step-desc">浏览器验证：Playwright 重新操作确认修复</span></div>
+    <div class="flow-step"><span class="step-tag red">Step 6</span><span class="step-desc">回归测试：确保无新问题引入</span></div>
+    <div class="flow-step"><span class="step-tag accent">Step 7</span><span class="step-desc">修复报告：根因+修复+验证结果</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>remediation-expert / browser-test-expert</div>
+</div>
+</div>
+<!-- 测试体系 -->
+<h3>六、测试体系（6条）</h3>
+<div class="cmd-grid">
+<div class="cmd-card">
+  <div class="cmd-name">/test-unit</div>
+  <div class="cmd-desc"><strong>单元测试</strong> — TDD 循环：检测框架→分析代码→生成(Red)→运行(Green)→重构(Refactor)。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Step 1</span><span class="step-desc">框架检测：识别项目测试框架（vitest/jest/mocha）</span></div>
+    <div class="flow-step"><span class="step-tag green">Step 2</span><span class="step-desc">代码分析：识别测试目标（函数/类/模块）</span></div>
+    <div class="flow-step"><span class="step-tag red">Step 3</span><span class="step-desc">Red：生成失败测试（验证测试有效性）</span></div>
+    <div class="flow-step"><span class="step-tag green">Step 4</span><span class="step-desc">Green：生成最小实现使测试通过</span></div>
+    <div class="flow-step"><span class="step-tag orange">Step 5</span><span class="step-desc">Refactor：重构代码，保持测试通过</span></div>
+  </div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/test-integration</div>
+  <div class="cmd-desc"><strong>集成测试</strong> — 识别 API 契约→启动环境→生成测试→运行→清理环境。验证模块间交互。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Step 1</span><span class="step-desc">契约识别：读取 API 定义（OpenAPI/GraphQL schema/路由）</span></div>
+    <div class="flow-step"><span class="step-tag orange">Step 2</span><span class="step-desc">环境准备：启动测试数据库/模拟服务</span></div>
+    <div class="flow-step"><span class="step-tag accent">Step 3</span><span class="step-desc">测试生成：按契约生成请求/响应验证用例</span></div>
+    <div class="flow-step"><span class="step-tag green">Step 4</span><span class="step-desc">执行+清理：运行测试→清理环境</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>api-contract-expert / api-test-expert</div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/test-e2e</div>
+  <div class="cmd-desc"><strong>端到端测试</strong> — 用户故事→选择工具→编写用例→运行→报告。覆盖完整用户流程。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Step 1</span><span class="step-desc">用户故事提取：从需求文档提取关键用户旅程</span></div>
+    <div class="flow-step"><span class="step-tag orange">Step 2</span><span class="step-desc">工具选择：Playwright/Cypress/Selenium 按项目适配</span></div>
+    <div class="flow-step"><span class="step-tag accent">Step 3</span><span class="step-desc">用例编写：按用户旅程编写 E2E 测试脚本</span></div>
+    <div class="flow-step"><span class="step-tag green">Step 4</span><span class="step-desc">运行+报告：执行全部 E2E → 通过/失败清单 + 截图</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>e2e-test-expert / browser-test-expert</div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/test-perf</div>
+  <div class="cmd-desc"><strong>性能测试</strong> — 定义目标→选择工具→建立基线→负载/压力测试→定位瓶颈→优化建议报告。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Step 1</span><span class="step-desc">目标定义：RT/吞吐量/并发数/资源使用指标</span></div>
+    <div class="flow-step"><span class="step-tag orange">Step 2</span><span class="step-desc">工具选择：k6/autocannon/artillery 按场景适配</span></div>
+    <div class="flow-step"><span class="step-tag accent">Step 3</span><span class="step-desc">基线+负载：建立性能基线→逐步加压→记录拐点</span></div>
+    <div class="flow-step"><span class="step-tag red">Step 4</span><span class="step-desc">瓶颈定位+报告：火焰图/慢查询/内存分析→优化建议</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>perf-test-expert / perf-review-expert</div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/test-security</div>
+  <div class="cmd-desc"><strong>安全测试</strong> — 确认授权范围→爬取→主动漏洞扫描（OWASP Top 10）→修复验证→安全审计报告。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag red">Step 1</span><span class="step-desc">授权确认：明确扫描范围+授权边界（必须）</span></div>
+    <div class="flow-step"><span class="step-tag orange">Step 2</span><span class="step-desc">信息收集：爬取端点+识别技术栈+攻击面分析</span></div>
+    <div class="flow-step"><span class="step-tag red">Step 3</span><span class="step-desc">漏洞扫描：OWASP Top 10（注入/XSS/SSRF/越权等）</span></div>
+    <div class="flow-step"><span class="step-tag green">Step 4</span><span class="step-desc">修复验证+报告：确认修复→安全审计报告</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>security-review-expert</div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/browser-test</div>
+  <div class="cmd-desc"><strong>浏览器交互测试</strong> — 文档驱动闭环：test-doc-writer→test-executor→remediation-expert。最多 2 轮修复重测。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Step 1</span><span class="step-desc">用例编写：test-doc-writer 编写结构化测试文档（步骤+预期）</span></div>
+    <div class="flow-step"><span class="step-tag orange">Step 2</span><span class="step-desc">按文档执行：test-executor 严格按文档执行，不过度发挥</span></div>
+    <div class="flow-step"><span class="step-tag red">Step 3</span><span class="step-desc">失败修复：remediation-expert 分析失败→spawn 修复→重测≤2轮</span></div>
+    <div class="flow-step"><span class="step-tag accent">Step 4</span><span class="step-desc">测试报告：通过/失败清单+截图+修复记录</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>test-doc-writer / test-executor / remediation-expert / remediation-planner</div>
+</div>
+</div>
+<!-- 专家咨询 -->
+<h3>七、专家咨询（3条）</h3>
+<div class="cmd-grid">
+<div class="cmd-card">
+  <div class="cmd-name">/frontend-architect</div>
+  <div class="cmd-desc"><strong>前端架构咨询</strong> — 只读模式，不修改文件。spawn frontend-architect Agent。适用：组件架构设计、性能优化方案、状态管理选型、微前端架构。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Step 1</span><span class="step-desc">问题收集：提取架构问题+当前技术栈+约束条件</span></div>
+    <div class="flow-step"><span class="step-tag orange">Step 2</span><span class="step-desc">spawn frontend-architect：只读分析+方案设计</span></div>
+    <div class="flow-step"><span class="step-tag accent">Step 3</span><span class="step-desc">输出呈现：架构方案+对比分析+推荐路径</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>frontend-architect</div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/backend-architect</div>
+  <div class="cmd-desc"><strong>后端架构咨询</strong> — 只读模式。spawn backend-architect Agent。适用：API 设计、数据库架构、服务拆分、消息队列选型。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Step 1</span><span class="step-desc">问题收集：提取架构问题+当前技术栈+约束条件</span></div>
+    <div class="flow-step"><span class="step-tag orange">Step 2</span><span class="step-desc">spawn backend-architect + database-architect：只读分析</span></div>
+    <div class="flow-step"><span class="step-tag accent">Step 3</span><span class="step-desc">输出呈现：架构方案+数据模型+推荐路径</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>backend-architect / database-architect</div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/algorithm-expert</div>
+  <div class="cmd-desc"><strong>算法设计与优化咨询</strong> — 只读模式。spawn algorithm-expert Agent。适用：复杂度优化、数据结构选型、数学建模、竞赛题目。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Step 1</span><span class="step-desc">问题收集：提取算法问题+约束+数据规模</span></div>
+    <div class="flow-step"><span class="step-tag orange">Step 2</span><span class="step-desc">spawn algorithm-expert：复杂度分析+多方案对比</span></div>
+    <div class="flow-step"><span class="step-tag accent">Step 3</span><span class="step-desc">输出呈现：最优算法+复杂度证明+实现伪代码</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>algorithm-expert</div>
+</div>
+</div>
+<!-- 发布与工具 -->
+<h3>八、发布与工具（4条）</h3>
+<div class="cmd-grid">
+<div class="cmd-card">
+  <div class="cmd-name">/release</div>
+  <div class="cmd-desc"><strong>快速发布</strong> — 5 Gate：环境检测→质量门→版本递增+CHANGELOG→commit+tag+push→验证(CI+npm)。适合个人/小团队快速发布。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Gate RL0</span><span class="step-desc">环境检测：检查工作区干净+分支正确+CI绿色</span></div>
+    <div class="flow-step"><span class="step-tag orange">Gate RL1</span><span class="step-desc">质量门：Lint+Type-check+Build+Test 全部通过</span></div>
+    <div class="flow-step"><span class="step-tag accent">Gate RL2</span><span class="step-desc">版本递增：patch/minor/major + CHANGELOG 更新</span></div>
+    <div class="flow-step"><span class="step-tag red">Gate RL3</span><span class="step-desc">发布：git commit+tag+push → CI 自动 npm publish</span></div>
+    <div class="flow-step"><span class="step-tag green">Gate RL4</span><span class="step-desc">验证：npm view + GitHub Release 确认</span></div>
+  </div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/publish</div>
+  <div class="cmd-desc"><strong>完整发布流程</strong>（含 PR+审查+合并）。工作分支→质量门→测试→版本→commit→push→PR→审查→合并→tag→发布。适合团队协作。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Step 1</span><span class="step-desc">工作分支：从 main 创建 release/x.y.z 分支</span></div>
+    <div class="flow-step"><span class="step-tag orange">Step 2</span><span class="step-desc">质量门+测试：全部通过</span></div>
+    <div class="flow-step"><span class="step-tag accent">Step 3</span><span class="step-desc">版本+commit+push：更新版本号→提交→推送</span></div>
+    <div class="flow-step"><span class="step-tag red">Step 4</span><span class="step-desc">PR+审查+合并：创建 PR→团队审查→合并到 main</span></div>
+    <div class="flow-step"><span class="step-tag green">Step 5</span><span class="step-desc">Tag+发布：打 tag→push→CI 自动发布</span></div>
+  </div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/sync</div>
+  <div class="cmd-desc"><strong>文档同步</strong> — 检查 CLAUDE.md/AGENTS.md/README.md/CHANGELOG.md 与代码一致性，清理过时文件。支持 --dry-run 预览模式。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Step 1</span><span class="step-desc">文档扫描：读取所有核心文档+代码统计</span></div>
+    <div class="flow-step"><span class="step-tag orange">Step 2</span><span class="step-desc">一致性检查：版本号/Agent数/Command数/Skill数交叉验证</span></div>
+    <div class="flow-step"><span class="step-tag red">Step 3</span><span class="step-desc">差异报告：列出不一致项+过时文件</span></div>
+    <div class="flow-step"><span class="step-tag accent">Step 4</span><span class="step-desc">自动修复：更新统计数字+清理过时文件（--dry-run 仅预览）</span></div>
+  </div>
+</div>
+<div class="cmd-card">
+  <div class="cmd-name">/browser-explore</div>
+  <div class="cmd-desc"><strong>浏览器自主探索</strong> — browser-use Agent 自主探索页面+发现 UI bug+产出结构化报告。适合 UI 验收测试。</div>
+  <div class="cmd-flow">
+    <div class="flow-step"><span class="step-tag green">Step 1</span><span class="step-desc">目标设定：定义探索范围+检查维度</span></div>
+    <div class="flow-step"><span class="step-tag orange">Step 2</span><span class="step-desc">自主探索：browser-use Agent 自动浏览+交互+截图</span></div>
+    <div class="flow-step"><span class="step-tag red">Step 3</span><span class="step-desc">Bug 发现：自动识别 UI 异常/布局问题/交互缺陷</span></div>
+    <div class="flow-step"><span class="step-tag accent">Step 4</span><span class="step-desc">结构化报告：发现清单+截图+严重度+复现步骤</span></div>
+  </div>
+  <div class="agents"><span>关键 Agent：</span>browser-use-expert</div>
+</div>
+</div>
+</div>
+<!-- ============================================================ -->
+<!-- 智能体体系 -->
+<!-- ============================================================ -->
+<h2 id="agents">智能体体系 <span class="count">— 71 个 Agent，6 大类别</span></h2>
+<div class="intro">
+  <p>每个 Agent 是高度专业化的 AI 角色，在流水线对应 Gate 被编排者 spawn 执行。Agent 读取 <code>AGENTS.md</code> 获取项目约束，遵循 Gate 操作权限表，产出结构化文档。</p>
+</div>
+<div class="agent-category">
+  <h4>&#x1F4DD; 实现类 <span class="count">25 个</span></h4>
+  <div class="agent-list">
+    <span class="agent-tag">frontend-dev-expert</span>
+    <span class="agent-tag">frontend-ui-expert</span>
+    <span class="agent-tag">frontend-state-expert</span>
+    <span class="agent-tag">backend-dev-expert</span>
+    <span class="agent-tag">backend-api-expert</span>
+    <span class="agent-tag">backend-logic-expert</span>
+    <span class="agent-tag">backend-data-expert</span>
+    <span class="agent-tag">android-dev-expert</span>
+    <span class="agent-tag">android-ui-expert</span>
+    <span class="agent-tag">android-state-expert</span>
+    <span class="agent-tag">ios-dev-expert</span>
+    <span class="agent-tag">ios-ui-expert</span>
+    <span class="agent-tag">ios-state-expert</span>
+    <span class="agent-tag">flutter-dev-expert</span>
+    <span class="agent-tag">flutter-ui-expert</span>
+    <span class="agent-tag">flutter-state-expert</span>
+    <span class="agent-tag">expo-dev-expert</span>
+    <span class="agent-tag">expo-ui-expert</span>
+    <span class="agent-tag">expo-state-expert</span>
+    <span class="agent-tag">react-native-dev-expert</span>
+    <span class="agent-tag">react-native-ui-expert</span>
+    <span class="agent-tag">react-native-state-expert</span>
+    <span class="agent-tag">taro-dev-expert</span>
+    <span class="agent-tag">taro-ui-expert</span>
+    <span class="agent-tag">taro-state-expert</span>
+  </div>
+</div>
+<div class="agent-category">
+  <h4>&#x1F9EA; 测试类 <span class="count">15 个</span></h4>
+  <div class="agent-list">
+    <span class="agent-tag">frontend-test-expert</span>
+    <span class="agent-tag">backend-test-expert</span>
+    <span class="agent-tag">android-test-expert</span>
+    <span class="agent-tag">ios-test-expert</span>
+    <span class="agent-tag">flutter-test-expert</span>
+    <span class="agent-tag">taro-test-expert</span>
+    <span class="agent-tag">expo-test-expert</span>
+    <span class="agent-tag">react-native-test-expert</span>
+    <span class="agent-tag">browser-test-expert</span>
+    <span class="agent-tag">browser-use-expert</span>
+    <span class="agent-tag">e2e-test-expert</span>
+    <span class="agent-tag">perf-test-expert</span>
+    <span class="agent-tag">api-test-expert</span>
+    <span class="agent-tag">test-doc-writer</span>
+    <span class="agent-tag">test-executor</span>
+  </div>
+</div>
+<div class="agent-category">
+  <h4>&#x1F4CB; 规划/任务类 <span class="count">4 个</span></h4>
+  <div class="agent-list">
+    <span class="agent-tag">task-design</span>
+    <span class="agent-tag">planner</span>
+    <span class="agent-tag">skill-assignment-expert</span>
+    <span class="agent-tag">remediation-planner</span>
+  </div>
+</div>
+<div class="agent-category">
+  <h4>&#x1F50D; 审查类 <span class="count">16 个</span></h4>
+  <div class="agent-list">
+    <span class="agent-tag">frontend-review-expert</span>
+    <span class="agent-tag">backend-review-expert</span>
+    <span class="agent-tag">android-review-expert</span>
+    <span class="agent-tag">ios-review-expert</span>
+    <span class="agent-tag">flutter-review-expert</span>
+    <span class="agent-tag">taro-review-expert</span>
+    <span class="agent-tag">expo-review-expert</span>
+    <span class="agent-tag">react-native-review-expert</span>
+    <span class="agent-tag">diff-review-expert</span>
+    <span class="agent-tag">project-review-expert</span>
+    <span class="agent-tag">perf-review-expert</span>
+    <span class="agent-tag">security-review-expert</span>
+    <span class="agent-tag">qa-review-expert</span>
+    <span class="agent-tag">change-review-expert</span>
+    <span class="agent-tag">review-only</span>
+    <span class="agent-tag">review-fix-optimize</span>
+  </div>
+</div>
+<div class="agent-category">
+  <h4>&#x1F3DB; 架构/专家类 <span class="count">4 个</span></h4>
+  <div class="agent-list">
+    <span class="agent-tag">frontend-architect</span>
+    <span class="agent-tag">backend-architect</span>
+    <span class="agent-tag">database-architect</span>
+    <span class="agent-tag">algorithm-expert</span>
+  </div>
+</div>
+<div class="agent-category">
+  <h4>&#x1F50C; 探索/支撑类 <span class="count">7 个</span></h4>
+  <div class="agent-list">
+    <span class="agent-tag">code-explore-expert</span>
+    <span class="agent-tag">external-resource-expert</span>
+    <span class="agent-tag">api-contract-expert</span>
+    <span class="agent-tag">docs-engineer</span>
+    <span class="agent-tag">infra-deploy-expert</span>
+    <span class="agent-tag">remediation-expert</span>
+    <span class="agent-tag">docs-research-expert</span>
+  </div>
+</div>
+<!-- ============================================================ -->
+<!-- 流水线架构 -->
+<!-- ============================================================ -->
+<h2 id="pipelines">流水线架构</h2>
+<div class="intro">
+  <p>每条指令对应一条流水线，每条流水线由一组 Gate 组成。Gate 是 FSM 硬约束检查点——<strong>前一个 Gate 未通过，无法推进到下一个</strong>。引擎通过 MCP 工具 <code>gate_check</code> 和 <code>gate_enforce</code> 在每次 Agent spawn 后自动检查 Gate 状态，禁止跳过或回退。</p>
+</div>
+<h3>标准流水线 (full/frontend/backend/lite)</h3>
+<table class="pipe-table">
+<thead>
+<tr><th>Gate</th><th>名称</th><th>产出</th><th>操作权限</th><th>调度策略</th></tr>
+</thead>
+<tbody>
+<tr>
+  <td class="pipe-name">Gate A</td>
+  <td>需求澄清</td>
+  <td>REQ-XXX 需求文档</td>
+  <td>read, write_doc</td>
+  <td><span class="tag tag-blue">subagent_only</span></td>
+</tr>
+<tr>
+  <td class="pipe-name">Gate B-DDD</td>
+  <td>领域驱动分析</td>
+  <td>聚合/实体/值对象/领域服务</td>
+  <td>read, write_doc</td>
+  <td><span class="tag tag-blue">subagent_only</span></td>
+</tr>
+<tr>
+  <td class="pipe-name">Gate B-BDD</td>
+  <td>行为驱动场景</td>
+  <td>Gherkin 验收场景（条件性）</td>
+  <td>read, write_doc</td>
+  <td><span class="tag tag-blue">subagent_only</span></td>
+</tr>
+<tr>
+  <td class="pipe-name">Gate B-TDD</td>
+  <td>测试驱动分解</td>
+  <td>TASK-XXX → REQ-XXX 映射</td>
+  <td>read, write_doc</td>
+  <td><span class="tag tag-blue">subagent_only</span></td>
+</tr>
+<tr>
+  <td class="pipe-name">Gate B1</td>
+  <td>架构评审</td>
+  <td>架构评审报告（条件性）</td>
+  <td>read, write_doc, sweep_arch</td>
+  <td><span class="tag tag-blue">subagent_only</span></td>
+</tr>
+<tr>
+  <td class="pipe-name">Gate C</td>
+  <td>执行规划</td>
+  <td>parallel_batches 分组</td>
+  <td>read, write_doc, spawn_impl</td>
+  <td><span class="tag tag-blue">subagent_only</span></td>
+</tr>
+<tr>
+  <td class="pipe-name">Gate C-impl</td>
+  <td>并行实现</td>
+  <td>代码实现 + 自查报告</td>
+  <td>read, write_code, spawn_impl</td>
+  <td><span class="tag tag-green">prefer_team</span></td>
+</tr>
+<tr>
+  <td class="pipe-name">Gate C1</td>
+  <td>代码质量门</td>
+  <td>Lint+Type-check+Build+Deps Audit</td>
+  <td>read, lint, build, fix</td>
+  <td><span class="tag tag-blue">subagent_only</span></td>
+</tr>
+<tr>
+  <td class="pipe-name">Gate C1.5</td>
+  <td>视觉验证</td>
+  <td>截图对比报告（条件性）</td>
+  <td>read, preview, fix</td>
+  <td><span class="tag tag-blue">subagent_only</span></td>
+</tr>
+<tr>
+  <td class="pipe-name">Gate C2</td>
+  <td>测试验证</td>
+  <td>测试报告 · 最多5次重试</td>
+  <td>read, spawn_test, fix</td>
+  <td><span class="tag tag-green">prefer_team</span></td>
+</tr>
+<tr>
+  <td class="pipe-name">Gate D</td>
+  <td>评审</td>
+  <td>审查报告+严重度分级</td>
+  <td>read, review, audit, fix</td>
+  <td><span class="tag tag-green">prefer_team</span></td>
+</tr>
+<tr>
+  <td class="pipe-name">Gate E</td>
+  <td>发布上线</td>
+  <td>版本递增+CHANGELOG+发布</td>
+  <td>read, deploy, write_doc</td>
+  <td><span class="tag tag-blue">subagent_only</span></td>
+</tr>
+</tbody>
+</table>
+<h3>专业流水线（13条）</h3>
+<table class="pipe-table">
+<thead>
+<tr><th>流水线</th><th>指令</th><th>Gate 序列</th><th>门数</th><th>核心特色</th></tr>
+</thead>
+<tbody>
+<tr><td class="pipe-name">refactor</td><td><code>/refactor</code></td><td>R1→R2→R3→R4→R5</td><td>5</td><td>基线测试+行为漂移检测+自动回滚</td></tr>
+<tr><td class="pipe-name">hotfix</td><td><code>/hotfix</code></td><td>H0→H1→H2→H3</td><td>4</td><td>紧急审批+回滚预案+事后审计</td></tr>
+<tr><td class="pipe-name">migrate</td><td><code>/migrate</code></td><td>M1→M2→M3→M4</td><td>4</td><td>自动修复Lint≤2轮</td></tr>
+<tr><td class="pipe-name">evaluate</td><td><code>/evaluate</code></td><td>E0→E1→E2→E3</td><td>4</td><td>原型+量化指标+雷达图</td></tr>
+<tr><td class="pipe-name">debug</td><td><code>/debug</code></td><td>D0→D1→D2→D3→D4</td><td>5</td><td>最小复现+交互式诊断</td></tr>
+<tr><td class="pipe-name">research</td><td><code>/research</code></td><td>RS0→RS1→RS2→RS3→RS4</td><td>5</td><td>竞争假设+证据矩阵+假设验证</td></tr>
+<tr><td class="pipe-name">release</td><td><code>/release</code></td><td>RL0→RL1→RL2→RL3→RL4</td><td>5</td><td>质量门前置+CI自动发布</td></tr>
+<tr><td class="pipe-name">ask</td><td><code>/ask</code></td><td>K0→K1→K2→K3</td><td>4</td><td>4模式自适应+5维评分矩阵</td></tr>
+<tr><td class="pipe-name">simplify</td><td><code>/simplify</code></td><td>S0→S1→S2→S3</td><td>4</td><td>五桶分析+失败自动回滚</td></tr>
+<tr><td class="pipe-name">trace</td><td><code>/trace</code></td><td>T0→T1→T2→T3→T4</td><td>5</td><td>贝叶斯概率更新+>70%置信度</td></tr>
+<tr><td class="pipe-name">improve</td><td><code>/improve</code></td><td>IM0→IM1→IM2→IM3→IM4</td><td>5</td><td>benchmark迭代循环(allow_jump)</td></tr>
+</tbody>
+</table>
+<!-- ============================================================ -->
+<!-- 技能体系 -->
+<!-- ============================================================ -->
+<h2 id="skills">技能体系 <span class="count">— 34 个 Skill，12 个领域</span></h2>
+<table class="pipe-table">
+<thead>
+<tr><th>领域</th><th>技能</th></tr>
+</thead>
+<tbody>
+<tr><td><strong>基础</strong></td><td>behavioral-guidelines · context-engineering · using-agent-skills</td></tr>
+<tr><td><strong>需求</strong></td><td>spec-driven-development · idea-refine</td></tr>
+<tr><td><strong>规划</strong></td><td>planning-and-task-breakdown</td></tr>
+<tr><td><strong>实现</strong></td><td>source-driven-development · incremental-implementation · test-driven-development · code-standards · code-simplification · frontend-design · refactoring</td></tr>
+<tr><td><strong>质量</strong></td><td>code-quality-gate · code-review-and-quality · verification-before-completion</td></tr>
+<tr><td><strong>测试</strong></td><td>perf-testing · test-data-factory</td></tr>
+<tr><td><strong>调试</strong></td><td>debugging-and-error-recovery · debugging-deep</td></tr>
+<tr><td><strong>浏览器</strong></td><td>agent-browser · browser-testing · browser-use</td></tr>
+<tr><td><strong>安全</strong></td><td>security-and-hardening · security-testing</td></tr>
+<tr><td><strong>流程</strong></td><td>shipping-and-launch · git-workflow-and-versioning · finishing-a-development-branch</td></tr>
+<tr><td><strong>文档</strong></td><td>chinese-documentation · documentation-and-adrs · writing-skills</td></tr>
+<tr><td><strong>工具</strong></td><td>find-docs · find-skills · mcp-builder</td></tr>
+</tbody>
+</table>
+<!-- ============================================================ -->
+<!-- 使用推荐 -->
+<!-- ============================================================ -->
+<h2 id="scenarios">使用推荐</h2>
+<div class="recommend">
+  <strong>&#x1F4A1; 新手入门路径：</strong><br>
+  <code>/auto 需求描述</code> → 自动路由 + 引导 →
+  <code>/review</code> 审查产出 →
+  <code>/review-fix</code> 修复问题 →
+  <code>/release</code> 发布<br>
+  <strong>不需要记住所有指令，99% 的情况 /auto 就够了。</strong>
+</div>
+<div class="scenario-grid">
+  <div class="scenario-card">
+    <h4>&#x1F4BB; 日常开发（默认）</h4>
+    <p>使用 <code>/auto</code>，自动检测任务类型→路由最优流水线→跳过无关Gate→按复杂度分配Agent。最省心。</p>
+  </div>
+  <div class="scenario-card">
+    <h4>&#x1F3D7; 中大型功能</h4>
+    <p>使用 <code>/jarvis</code>，13 Gate 完整生命周期。DDD→BDD→TDD→架构→规划→并行实现→视觉验证→测试→评审→发布，质量保证最强。</p>
+  </div>
+  <div class="scenario-card">
+    <h4>&#x1F527; 重构代码</h4>
+    <p>使用 <code>/refactor</code>，R1-R5 安全网保护。建立基线测试→执行重构→行为漂移检测→失败自动回滚。</p>
+  </div>
+  <div class="scenario-card">
+    <h4>&#x1F6A8; 紧急修复线上故障</h4>
+    <p>使用 <code>/hotfix</code>，H0-H3 紧急协议。影响声明+审批→最小化修复→快速验证+回滚预案→事后审计。</p>
+  </div>
+  <div class="scenario-card">
+    <h4>&#x1F9F9; 代码质量清理</h4>
+    <p>使用 <code>/simplify</code>，五桶分析（死代码/重复/过度抽象/AI痕迹/复杂性）。失败自动回滚，不怕改坏。</p>
+  </div>
+  <div class="scenario-card">
+    <h4>&#x1F50D; 排查 Bug</h4>
+    <p>简单问题用 <code>/debug</code>（交互式诊断），复杂根因用 <code>/trace</code>（贝叶斯因果推理，多假设竞争+证据更新）。</p>
+  </div>
+  <div class="scenario-card">
+    <h4>&#x26A1; 性能优化</h4>
+    <p>使用 <code>/improve</code>，度量驱动迭代。量化目标→研究→计划→执行+benchmark→评估→不达标继续迭代。</p>
+  </div>
+  <div class="scenario-card">
+    <h4>&#x1F4CA; 技术选型</h4>
+    <p>快速选型用 <code>/evaluate</code>（原型+量化指标+雷达图），深度调研用 <code>/research</code>（5阶段分析+竞争假设验证）。</p>
+  </div>
+  <div class="scenario-card">
+    <h4>&#x2753; 需求不清晰</h4>
+    <p>使用 <code>/ask</code>（Interview 模式），Socratic 追问→K1探索空间→K2五维评分矩阵→K3产出分析报告+路由建议。</p>
+  </div>
+  <div class="scenario-card">
+    <h4>&#x1F4E6; 框架升级</h4>
+    <p>使用 <code>/migrate</code>，M1-M4 迁移流程。定义规则→应用迁移→编译验证→自动修复Lint≤2轮。</p>
+  </div>
+  <div class="scenario-card">
+    <h4>&#x1F4DD; 代码审查</h4>
+    <p>使用 <code>/review</code>（只读审查）→ <code>/review-fix</code>（修复闭环）。先审查后修复再复审，最多2轮。</p>
+  </div>
+  <div class="scenario-card">
+    <h4>&#x1F4E4; 发布上线</h4>
+    <p>个人/小团队用 <code>/release</code>（质量门→版本递增→CI自动发布），团队协作用 <code>/publish</code>（含 PR+审查+合并流程）。</p>
+  </div>
+  <div class="scenario-card">
+    <h4>&#x1F3AF; 前端开发</h4>
+    <p>使用 <code>/frontend</code>，C1.5 视觉验证强制。spawn frontend-dev/ui/state-expert 并行实现，frontend-architect 架构评审。</p>
+  </div>
+  <div class="scenario-card">
+    <h4>&#x1F310; 后端开发</h4>
+    <p>使用 <code>/backend</code>，跳过 C1.5 视觉验证。spawn backend-dev/api/logic/data-expert 并行实现，database-architect 数据模型评审。</p>
+  </div>
+  <div class="scenario-card">
+    <h4>&#x1F4F1; 移动端/跨端开发</h4>
+    <p>使用平台对应指令（<code>/android</code> <code>/ios</code> <code>/flutter</code> <code>/expo</code> <code>/react-native</code> <code>/taro</code>），每个平台有专属 dev/ui/state/test/review 全套 Agent。</p>
+  </div>
+  <div class="scenario-card">
+    <h4>&#x1F9EA; 测试补齐</h4>
+    <p>按需选择：<code>/test-unit</code>（单元）<code>/test-integration</code>（集成）<code>/test-e2e</code>（端到端）<code>/test-perf</code>（性能）<code>/test-security</code>（安全）。前端用 <code>/browser-test</code> 文档驱动测试。</p>
+  </div>
+</div>
+<!-- ============================================================ -->
+<!-- 快速开始 -->
+<!-- ============================================================ -->
+<h2 id="quickstart">快速开始</h2>
+<div class="intro">
+  <p><strong>3 步开始：</strong></p>
+  <p>1. <code>npm i -g jarvis-agent-factory</code> — 安装 CLI（零原生依赖，Node 22.5+ 内置 node:sqlite）</p>
+  <p>2. <code>jarvis init -y</code> — 一键部署配置 + MCP + 钩子到当前项目</p>
+  <p>3. 重启 Claude Code → 引擎自动拉起 → 输入 <code>/auto 你的任务</code> 开始使用</p>
+  <br>
+  <p><strong>Web 面板：</strong><code>jarvis web</code> → <code>http://localhost:3457</code> 查看流水线看板、会话管理、Agent 配置、归档记录。</p>
+  <p><strong>远程面板：</strong>每次 GitHub Release 附带单 HTML 文件，下载即可打开，无需本地安装。</p>
+</div>
+<!-- ============================================================ -->
+<!-- FOOTER -->
+<!-- ============================================================ -->
+<div class="footer">
+  <p>
+    <strong>Jarvis Agent Factory v4.2.1</strong> &nbsp;·&nbsp;
+    <a href="https://github.com/Wjl1224734792/Jarvis-Agent-Factory">GitHub</a> &nbsp;·&nbsp;
+    <a href="https://www.npmjs.com/package/jarvis-agent-factory">npm</a> &nbsp;·&nbsp;
+    <a href="https://github.com/Wjl1224734792/Jarvis-Agent-Factory/blob/main/AGENTS.md">AGENTS.md</a> &nbsp;·&nbsp;
+    <a href="https://github.com/Wjl1224734792/Jarvis-Agent-Factory/blob/main/README.md">README.md</a>
+  </p>
+  <p style="margin-top:8px">39 Commands · 71 Agents · 34 Skills · 15 Pipelines · 62 Gates · MIT License</p>
+</div>
+</body>
+</html>

package/package.json CHANGED Viewed

@@ -1,6 +1,6 @@
 {
   "name": "jarvis-agent-factory",
-  "version": "4.2.1",
+  "version": "4.2.2",
   "description": "Jarvis Agent Factory CLI — Claude Code 多智能体 AI 编程助手配置安装器 | Multi-agent AI coding assistant config installer for Claude Code",
   "keywords": [
     "jarvis",