openmatrix 0.2.20 → 0.2.21

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "openmatrix",
-  "version": "0.2.20",
+  "version": "0.2.21",
   "description": "AI Agent task orchestration system with Claude Code Skills integration",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",

package/skills/approve.md

@@ -1,8 +1,39 @@
 ---
 name: om:approve
-description: "Use when handling pending approvals including plan review, merge confirmation, deploy approval, and blocked task decisions during OpenMatrix execution. Triggers on: 审批, approve, 批准, plan review, merge conflict resolution, deploy confirmation, 阻塞处理, technical decision, pending approval, waiting for approval, 待确认."
+description: "Use when handling pending approvals including plan review, merge confirmation, deploy approval, and blocked task decisions during OpenMatrix execution. Triggers on APPROVAL intent: user wants to review and approve pending items, handle merge conflicts, or confirm deploy. DO NOT trigger on: starting tasks, debugging, or status checks. Intent signals: user mentions 'approve', 'review plan', 'confirm merge', 'approve deploy', or refers to pending approval actions."
 ---
 
+<INTENT-JUDGMENT>
+## 意图判断指南
+
+**AI 应根据用户语义判断意图：**
+
+### 触发信号（审批意图）
+
+- 用户要审批计划
+- 处理合并冲突
+- 确认部署
+- 处理待审批项
+
+### 不触发信号
+
+| 用户意图 | 应调用 |
+|---------|--------|
+| 开始任务 | /om:start |
+| 调查问题 | /om:debug |
+| 查看状态 | /om:status |
+
+### 示例判断
+
+| 用户消息 | 判断 | 结果 |
+|---------|------|------|
+| "审批这个计划" | 审批意图 | 触发 ✓ |
+| "确认合并" | 确认意图 | 触发 ✓ |
+| "批准部署" | 部署审批意图 | 触发 ✓ |
+| "为什么没通过审批" | 调查意图 | /om:debug |
+| "开始实现" | 开发意图 | /om:start |
+</INTENT-JUDGMENT>
+
 <NO-OTHER-SKILLS>
 执行此技能时，不得调用 superpowers、gsd 或其他任务编排相关的技能。OpenMatrix 独立运行，不依赖外部任务编排系统。
 </NO-OTHER-SKILLS>

package/skills/auto.md

@@ -1,8 +1,39 @@
 ---
 name: om:auto
-description: "Use when the user wants fully automated task execution with zero manual approvals. Triggers on: 全自动, 无人值守, hands-free, non-stop, don't ask me, 直接执行, skip all confirmations, batch refactor, large migration, bulk changes. Use for multi-task execution where the user doesn't want to be interrupted at any approval point (plan/merge/deploy)."
+description: "Use when the user wants fully automated task execution with zero manual approvals. Triggers on AUTOMATION intent: user explicitly requests hands-free execution, wants to skip all confirmations, or needs batch processing without interruptions. DO NOT trigger on: interactive requests, debugging, or status checks. Intent signals: user says 'fully auto', 'hands-free', 'no interruptions', 'skip approvals', or describes large batch operations."
 ---
 
+<INTENT-JUDGMENT>
+## 意图判断指南
+
+**AI 应根据用户语义判断意图：**
+
+### 触发信号（全自动意图）
+
+- 用户明确要"全自动执行"
+- 需要跳过所有审批确认
+- 批量操作，不希望中断
+- 用户表达"直接执行"、"不要问我"
+
+### 不触发信号
+
+| 用户意图 | 应调用 |
+|---------|--------|
+| 需要交互确认 | /om:start |
+| 调查问题 | /om:debug |
+| 查看状态 | /om:status |
+
+### 示例判断
+
+| 用户消息 | 判断 | 结果 |
+|---------|------|------|
+| "全自动执行" | 全自动意图 | 触发 ✓ |
+| "不要问我直接做" | 无中断意图 | 触发 ✓ |
+| "批量重构跳过确认" | 批量自动化 | 触发 ✓ |
+| "交互式执行" | 需要确认 | /om:start |
+| "查看执行状态" | 状态检查 | /om:status |
+</INTENT-JUDGMENT>
+
 <NOTE>
 ## 注意：区分 `/om:auto` 指令与「全自动执行」模式
 

package/skills/brainstorm.md

@@ -1,8 +1,40 @@
 ---
 name: om:brainstorm
-description: "Use when the user wants to explore requirements, design alternatives, or validate ideas before implementation. Triggers on: 头脑风暴, 设计方案, 需求分析, 技术选型, architecture design, 'how should I build', multi-module features, new projects from scratch, unclear requirements, complex system design. Use even if the user just says 'implement X' and the task involves multiple modules or unclear requirements — brainstorm first, don't jump straight to coding."
+description: "Use when the user wants to explore requirements, design alternatives, or validate ideas before implementation. Triggers on CLARIFICATION/DESIGN intent: user needs to clarify unclear requirements, explore multiple approaches, or design architecture before coding. DO NOT trigger on: simple implementation requests (clear path), status checks, or pure information queries. Intent signals: user says 'how should I', mentions multiple options/uncertainties, describes complex system from scratch, or task scope is ambiguous."
 ---
 
+<INTENT-JUDGMENT>
+## 意图判断指南
+
+**AI 应根据用户语义判断意图：**
+
+### 触发信号（澄清/设计意图）
+
+- 用户想探索多种实现方案
+- 需求不明确，需要澄清
+- 涉及多模块协同，需要设计
+- 从零开始搭建，需要架构规划
+- 用户表达"怎么设计"、"什么方案"
+
+### 不触发信号
+
+| 用户意图 | 应调用 |
+|---------|--------|
+| 明确的实现任务 | /om:start 或 /om:feature |
+| 状态检查 | /om:status |
+| 简单问题咨询 | 直接回答 |
+
+### 示例判断
+
+| 用户消息 | 判断 | 结果 |
+|---------|------|------|
+| "登录功能怎么设计？" | 设计意图 | 触发 ✓ |
+| "从零搭建后台系统" | 架构规划意图 | 触发 ✓ |
+| "OAuth 选哪个方案？" | 方案探索意图 | 触发 ✓ |
+| "给按钮加点击事件" | 明确实现 | /om:feature |
+| "查看当前任务状态" | 状态检查 | /om:status |
+</INTENT-JUDGMENT>
+
 <NO-OTHER-SKILLS>
 **绝对禁止**调用以下技能（OpenMatrix 完全替代它们）:
 - ❌ superpowers:brainstorming → 你已经在 om:brainstorm 中了

package/skills/debug.md

@@ -1,6 +1,6 @@
 ---
 name: om:debug
-description: "Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes. Use for ANY technical issue: 测试失败, bug, 异常行为, 构建失败, 集成问题, 没触发, 没执行, 不工作, 原因, 为什么. Use ESPECIALLY when: under time pressure, 'just one quick fix' seems obvious, previous fix didn't work, you don't fully understand the issue. Don't skip when: issue seems simple (simple bugs have root causes too), you're in a hurry (systematic is faster than guessing)."
+description: "Use when user encounters ANY technical problem and needs systematic diagnosis. Triggers on PROBLEM-EXPLORATION intent: user wants to understand WHY something failed, investigate root cause, or diagnose unexpected behavior. DO NOT trigger on: simple questions (how to do X), implementation requests (build/fix/add), or status checks. Intent signals: user seems confused about failure, asks '为什么', mentions error/issue without clear solution, wants to investigate before fixing."
 priority: high
 ---
 
@@ -14,6 +14,61 @@ priority: high
 **诊断和修复阶段只能使用 Agent 工具** — 直接调用 Agent，不通过任何中间层。
 </NO-OTHER-SKILLS>
 
+<INTENT-JUDGMENT>
+## 意图判断指南
+
+**AI 应根据用户语义判断意图，而非关键词匹配：**
+
+### 触发信号（问题探索意图）
+
+用户表现出：
+- 需要理解失败原因（不知道为什么出错）
+- 想先调查再行动（排查、诊断、定位）
+- 对问题感到困惑（不清楚根因）
+- 多次尝试未成功（之前修复没生效）
+- 系统行为不符合预期（奇怪、意外）
+
+### 不触发信号（其他意图）
+
+| 用户意图 | 应调用的 skill |
+|---------|---------------|
+| 想要实现新功能 | /om 或 /om:start |
+| 想要修复已知问题（明确知道改什么） | /om:start |
+| 询问如何做某事 | 直接回答，不调用 skill |
+| 查看当前状态 | /om:status |
+| 简单问题咨询 | 直接回答 |
+
+### 判断流程
+
+```
+用户消息
+    │
+    ├── 是否描述了技术问题/失败？
+    │   ├─ 否 → 不触发
+    │   └─ 是 ↓
+    │
+    ├── 用户是否理解问题原因？
+    │   ├─ 是，明确知道怎么修 → /om:start (直接修复)
+    │   └─ 否，想先调查 → 触发 /om:debug ✓
+    │
+    ├── 用户是否想先诊断再修复？
+    │   ├─ 是 → 触发 /om:debug ✓
+    │   └─ 否 → 根据其他意图选择
+```
+
+### 示例判断
+
+| 用户消息 | 判断 | 结果 |
+|---------|------|------|
+| "测试失败了" | 描述问题，未说明原因 | 触发 debug |
+| "查一下为什么没触发" | 明确要调查 | 触发 debug ✓ |
+| "帮我排查这个问题" | 排查意图明确 | 触发 debug ✓ |
+| "不知道为什么出错了" | 困惑，需要诊断 | 触发 debug ✓ |
+| "修复登录页bug" | 明确要修复，知道改哪 | /om:start |
+| "API返回500，帮我改一下" | 要修复而非调查 | /om:start |
+| "怎么调试node进程" | 询问方法，非排查 | 直接回答 |
+</INTENT-JUDGMENT>
+
 <MANDATORY-EXECUTION-ORDER>
 ## 执行顺序 - 必须严格按此顺序，不得跳过
 

package/skills/deploy.md

@@ -1,8 +1,40 @@
 ---
 name: om:deploy
-description: "智能部署助手：分析项目+系统环境 → 推荐最可行方案 → 执行部署 → 生成一键脚本。Triggers on: deploy, 部署, 发布, docker, kubernetes, 环境搭建, 开发环境, make, taskfile, pm2"
+description: "Triggers on DEPLOYMENT intent: user wants to deploy, publish, set up environments, or create deployment scripts. DO NOT trigger on: development tasks, status checks, or debugging."
 ---
 
+<INTENT-JUDGMENT>
+## 意图判断指南
+
+**AI 应根据用户语义判断意图：**
+
+### 触发信号（部署意图）
+
+- 用户想部署/发布项目
+- 需要配置 Docker/Kubernetes
+- 环境搭建需求
+- 生成部署脚本
+- 发布到生产环境
+
+### 不触发信号
+
+| 用户意图 | 应调用 |
+|---------|--------|
+| 实现功能 | /om:start |
+| 修复 bug | /om:debug 或 /om:start |
+| 查看状态 | /om:status |
+
+### 示例判断
+
+| 用户消息 | 判断 | 结果 |
+|---------|------|------|
+| "部署到服务器" | 部署意图 | 触发 ✓ |
+| "用 Docker 运行" | 环境配置意图 | 触发 ✓ |
+| "发布 npm 包" | 发布意图 | 触发 ✓ |
+| "实现 API 接口" | 开发意图 | /om:start |
+| "为什么部署失败" | 调查意图 | /om:debug |
+</INTENT-JUDGMENT>
+
 <NO-OTHER-SKILLS>
 执行此技能时，不得调用 superpowers、gsd 或其他任务编排相关的技能。
 </NO-OTHER-SKILLS>
@@ -157,13 +189,11 @@ AskUserQuestion:
 ```bash
 docker build -t <project-name> .
 docker run -d -p <port>:<port> --name <project-name> <project-name>
-docker ps | grep <project-name>
 ```
 
 **Docker Compose：**
 ```bash
 docker-compose up -d --build
-docker-compose ps
 ```
 
 **Make：**
@@ -174,17 +204,133 @@ make deploy
 **npm scripts：**
 ```bash
 npm run build
-npm run start
+npm run start &
+```
+
+---
+
+## Step 7.1: 自动验证部署结果
+
+执行部署后，AI 自动执行以下验证（按顺序，全部通过才算成功）：
+
+**1. 检查容器/进程状态：**
+```bash
+# Docker
+docker ps --format "{{.Names}}\t{{.Status}}" | grep <project-name>
+
+# Docker Compose
+docker-compose ps
+
+# PM2
+pm2 list
+
+# 进程检查
+ps aux | grep <project-name>
 ```
 
-执行后验证：
+**2. 检查端口是否在监听：**
 ```bash
-# 检查服务是否正常运行
-docker ps 2>/dev/null | grep <name>
-curl -s http://localhost:<port>/health 2>/dev/null || echo "服务已启动"
+# Linux/Mac
+lsof -i :<port> 2>/dev/null || netstat -tlnp 2>/dev/null | grep <port>
+
+# Windows
+netstat -ano | findstr :<port>
+```
+
+**3. HTTP 连通性验证（核心验证）：**
+```bash
+# 等待服务启动（最多重试 5 次，每次间隔 2 秒）
+for i in 1 2 3 4 5; do
+  STATUS=$(curl -s -o /dev/null -w "%{http_code}" http://localhost:<port>/ 2>/dev/null)
+  if [ "$STATUS" -ge 200 ] && [ "$STATUS" -lt 400 ]; then
+    echo "HTTP $STATUS OK"
+    break
+  fi
+  sleep 2
+done
+
+# 获取响应内容摘要
+curl -s http://localhost:<port>/ 2>/dev/null | head -20
+```
+
+**4. 健康检查端点（如果存在）：**
+```bash
+curl -s http://localhost:<port>/health 2>/dev/null
+curl -s http://localhost:<port>/api/health 2>/dev/null
+curl -s http://localhost:<port>/healthz 2>/dev/null
+```
+
+**5. 浏览器可访问性验证（Web 项目）：**
+
+如果是 Web 项目，使用 Playwright 工具打开浏览器验证页面是否正常加载：
+```
+browser_navigate: http://localhost:<port>/
+browser_snapshot: 检查页面是否正常渲染
+browser_take_screenshot: 截图保存验证结果
+browser_close
+```
+
+如果无法使用浏览器工具，输出可访问的 URL 供用户手动验证：
+```
+🌐 请在浏览器中访问: http://localhost:<port>/
+```
+
+**6. 日志检查（任何验证失败时）：**
+```bash
+# Docker 日志
+docker logs <project-name> --tail 50
+
+# PM2 日志
+pm2 logs <project-name> --lines 50
+```
+
+---
+
+## Step 7.2: 输出验证报告
+
+AI 根据验证结果输出状态报告：
+
+**验证成功：**
+```markdown
+## ✅ 部署验证报告
+
+**服务状态**: 运行中 (Up 2 minutes)
+**端口**: 3000 已监听
+**HTTP 连通**: http://localhost:3000/ → 200 OK
+**健康检查**: /health → 200 OK
+
+### 验证清单
+- ✅ 容器/进程正常运行
+- ✅ 端口 3000 已监听
+- ✅ HTTP 请求返回 200
+- ✅ 页面可正常访问
+
+### 访问地址
+🌐 http://localhost:3000/
+
+### 快捷命令
+- 查看日志: docker logs -f <project-name>
+- 停止服务: docker stop <project-name>
+- 重启服务: docker restart <project-name>
+```
+
+**验证失败：**
+```markdown
+## ❌ 部署验证失败
+
+**问题**: 容器启动后立即退出
+**状态**: Exited (1)
+
+### 错误日志（最近 20 行）
+<显示 docker logs 输出>
+
+### 建议修复
+1. 检查 Dockerfile 配置
+2. 检查启动命令是否正确
+3. 查看完整日志: docker logs <project-name>
 ```
 
-输出执行结果，告知用户服务状态。
+验证失败时，询问用户是否要查看完整日志或尝试修复。
 
 ---
 

package/skills/feature.md

@@ -1,9 +1,40 @@
 ---
 name: om:feature
-description: "Use for small feature requests needing quick iteration without full task file management. Triggers on: 小需求, 小功能, 小改动, minor, quick, 快速, 简单, 添加按钮, 加个字段, 轻量. AI evaluates complexity and routes automatically from /om entry point."
+description: "Use for small feature requests needing quick iteration without full task file management. Triggers on MINOR-CHANGE intent: user wants small, quick changes with clear scope - single component, few files, straightforward implementation. DO NOT trigger on: complex multi-module tasks, unclear requirements, or investigation tasks. Intent signals: user explicitly says 'small/quick/simple', describes a focused change, or task is clearly localized."
 priority: high
 ---
 
+<INTENT-JUDGMENT>
+## 意图判断指南
+
+**AI 应根据用户语义判断意图：**
+
+### 触发信号（小需求意图）
+
+- 用户表达"小改动"、"快速"、"简单"
+- 任务范围明确且单一
+- 改动点集中在少数文件
+- 实现路径显而易见
+
+### 不触发信号
+
+| 用户意图 | 应调用 |
+|---------|--------|
+| 多模块复杂任务 | /om:start 或 /om:brainstorm |
+| 需求不明确 | /om:brainstorm |
+| 状态检查 | /om:status |
+
+### 示例判断
+
+| 用户消息 | 判断 | 结果 |
+|---------|------|------|
+| "给列表加搜索按钮" | 小改动意图 | 触发 ✓ |
+| "调整按钮颜色" | 简单修改意图 | 触发 ✓ |
+| "加个字段到表单" | 小需求意图 | 触发 ✓ |
+| "重构认证模块" | 复杂任务 | /om:start |
+| "从零搭建系统" | 需要设计 | /om:brainstorm |
+</INTENT-JUDGMENT>
+
 <NO-OTHER-SKILLS>
 **绝对禁止**调用以下技能：
 - ❌ superpowers:brainstorming → 用 /om:brainstorm 代替

package/skills/meeting.md

@@ -1,8 +1,39 @@
 ---
 name: om:meeting
-description: "Use when handling blocked tasks, technical decisions, or workflow interruptions during OpenMatrix execution. Triggers on: 阻塞, blocked, 决策, decision needed, 技术选型, database connection failed, API key missing, merge conflict, dependency issue, task cannot proceed. Use when the user reports something is stuck, waiting for info, or needs to make a choice that blocks execution."
+description: "Use when handling blocked tasks, technical decisions, or workflow interruptions during OpenMatrix execution. Triggers on BLOCKED/DECISION intent: user reports task is stuck, needs to make a blocking choice, or something prevents progress. DO NOT trigger on: starting tasks, debugging root causes, or status checks. Intent signals: user says 'blocked', 'stuck', 'waiting', 'cannot proceed', or describes a blocker/dependency issue."
 ---
 
+<INTENT-JUDGMENT>
+## 意图判断指南
+
+**AI 应根据用户语义判断意图：**
+
+### 触发信号（阻塞处理意图）
+
+- 任务被阻塞无法继续
+- 需要做技术决策
+- 等待外部信息
+- 执行中断
+
+### 不触发信号
+
+| 用户意图 | 应调用 |
+|---------|--------|
+| 开始新任务 | /om:start |
+| 调查失败原因 | /om:debug |
+| 查看进度 | /om:status |
+
+### 示例判断
+
+| 用户消息 | 判断 | 结果 |
+|---------|------|------|
+| "任务卡住了" | 阻塞意图 | 触发 ✓ |
+| "缺少 API key" | 阻塞问题 | 触发 ✓ |
+| "需要做技术决策" | 决策意图 | 触发 ✓ |
+| "为什么卡住" | 调查意图 | /om:debug |
+| "开始新功能" | 开发意图 | /om:start |
+</INTENT-JUDGMENT>
+
 <NO-OTHER-SKILLS>
 执行此技能时，不得调用 superpowers、gsd 或其他任务编排相关的技能。OpenMatrix 独立运行，不依赖外部任务编排系统。
 

package/skills/om.md

@@ -1,20 +1,50 @@
 ---
 name: om
-description: "Default entry point for all development tasks. Routes to /om:feature (small), /om:start (standard), or /om:brainstorm (clarify/design) automatically. Triggers on: implement, build, fix, create, refactor, 修复, 实现, 开发, 添加功能, new feature, 小需求, 小功能, 小改动, minor, quick, 快速, 简单, 添加按钮, 加个字段, 轻量. Use for ANY task that produces code changes — don't answer directly, route to OpenMatrix workflow."
+description: "Default entry point for ALL development tasks. Triggers on DEVELOPMENT intent: user wants to implement, build, fix, refactor, or create something that produces code changes. DO NOT trigger on: questions (how to?), information requests, status checks, or pure discussion. Intent signals: user describes an action that requires code modification, has a clear goal to add/change/remove functionality, or wants to execute a development task."
 priority: critical
 always_load: true
 ---
 
+<INTENT-JUDGMENT>
+## 意图判断指南
+
+**AI 应根据用户语义判断意图：**
+
+### 触发信号（开发意图）
+
+- 用户想实现/创建功能（产生代码变更）
+- 用户想修复已知问题（有明确的修复目标）
+- 用户想重构现有代码
+- 用户描述了具体的开发任务
+- 用户有明确的目标（添加/修改/删除功能）
+
+### 不触发信号（其他意图）
+
+| 用户意图 | 应调用 |
+|---------|--------|
+| 询问如何实现 | 直接回答或 /om:brainstorm |
+| 查看状态 | /om:status |
+| 调查问题原因 | /om:debug |
+| 纯讨论想法 | 直接交流或 /om:brainstorm |
+
+### 示例判断
+
+| 用户消息 | 判断 | 结果 |
+|---------|------|------|
+| "实现登录功能" | 开发意图明确 | 触发 ✓ |
+| "修复按钮样式问题" | 有明确修复目标 | 触发 ✓ |
+| "给列表加搜索" | 功能添加意图 | 触发 ✓ |
+| "怎么实现登录？" | 询问方法 | 直接回答 |
+| "API 返回 500 为什么" | 调查意图 | /om:debug |
+| "查看任务状态" | 状态检查 | /om:status |
+</INTENT-JUDGMENT>
+
 <EXTREMELY-IMPORTANT>
 ## This is the DEFAULT entry for ALL development tasks.
 
 **STOP. Do NOT invoke any other task orchestration skill.**
 **Invoke this skill IMMEDIATELY when the user wants code changes.**
-
-**Analyze user INTENT (decide in < 3 seconds):**
-- "实现X" / "implement X" / "build X" → Development task → This skill
-- "修复X" / "fix X" → Development task → This skill
-- "怎么实现?" / "how to?" → Question → NOT this skill (just answer)
+**意图判断参见上方 INTENT-JUDGMENT 区块。**
 </EXTREMELY-IMPORTANT>
 
 <NO-OTHER-SKILLS>

package/skills/report.md

@@ -1,8 +1,39 @@
 ---
 name: om:report
-description: "Use when generating a task execution report with statistics, task details, approval history, and agent performance. Triggers on: 报告, report, summary, 总结, execution stats, 统计, sprint report, weekly summary, 产出物, deliverables overview."
+description: "Use when generating a task execution report with statistics, task details, approval history, and agent performance. Triggers on REPORT intent: user wants a summary of execution, statistics overview, or deliverables documentation. DO NOT trigger on: starting tasks, debugging, or status checks. Intent signals: user asks for 'report', 'summary', 'statistics', 'what was done', or wants comprehensive output of execution results."
 ---
 
+<INTENT-JUDGMENT>
+## 意图判断指南
+
+**AI 应根据用户语义判断意图：**
+
+### 触发信号（报告意图）
+
+- 用户想要执行总结
+- 需要统计数据报告
+- 查看任务完成详情
+- 生成交付物概述
+
+### 不触发信号
+
+| 用户意图 | 应调用 |
+|---------|--------|
+| 查看实时进度 | /om:status |
+| 开始任务 | /om:start |
+| 调查问题 | /om:debug |
+
+### 示例判断
+
+| 用户消息 | 判断 | 结果 |
+|---------|------|------|
+| "生成报告" | 报告意图 | 触发 ✓ |
+| "执行总结" | 总结意图 | 触发 ✓ |
+| "查看统计" | 统计意图 | 触发 ✓ |
+| "当前进度" | 实时状态 | /om:status |
+| "为什么失败" | 调查意图 | /om:debug |
+</INTENT-JUDGMENT>
+
 <NO-OTHER-SKILLS>
 执行此技能时，不得调用 superpowers、gsd 或其他任务编排相关的技能。OpenMatrix 独立运行，不依赖外部任务编排系统。
 </NO-OTHER-SKILLS>

package/skills/research.md

@@ -1,8 +1,39 @@
 ---
 name: om:research
-description: "Use when conducting domain research before implementing vertical-domain tasks. Triggers on: 领域调研, research, game development, payment system, blockchain, AI application, 行业标准, tech stack exploration, domain analysis, unfamiliar vertical, need to understand the domain before building."
+description: "Use when conducting domain research before implementing vertical-domain tasks. Triggers on RESEARCH intent: user needs to understand an unfamiliar domain, explore tech stack, analyze industry standards, or gather knowledge before building. DO NOT trigger on: direct implementation, debugging, or status checks. Intent signals: user mentions 'research', 'understand domain', 'explore tech', or needs knowledge gathering before coding."
 ---
 
+<INTENT-JUDGMENT>
+## 意图判断指南
+
+**AI 应根据用户语义判断意图：**
+
+### 触发信号（调研意图）
+
+- 用户需要领域调研
+- 探索不熟悉的技术栈
+- 了解行业标准
+- 在实现前收集知识
+
+### 不触发信号
+
+| 用户意图 | 应调用 |
+|---------|--------|
+| 直接实现 | /om:start |
+| 需求澄清 | /om:brainstorm |
+| 查看状态 | /om:status |
+
+### 示例判断
+
+| 用户消息 | 判断 | 结果 |
+|---------|------|------|
+| "调研支付领域" | 调研意图 | 触发 ✓ |
+| "了解游戏开发" | 领域探索意图 | 触发 ✓ |
+| "研究区块链" | 技术调研意图 | 触发 ✓ |
+| "实现支付功能" | 实现意图 | /om:start |
+| "支付方案怎么设计" | 设计意图 | /om:brainstorm |
+</INTENT-JUDGMENT>
+
 <NO-OTHER-SKILLS>
 执行此技能时，不得调用其他任务编排相关的技能。OpenMatrix 独立运行，不依赖外部任务编排系统。
 

package/skills/resume.md

@@ -1,9 +1,39 @@
 ---
 name: om:resume
-description: "恢复中断的任务执行。智能检测数据源：轻量流程 (feature-session.json) 或完整流程 (state.json)。Triggers on: 恢复, resume, continue task, 继续执行, interrupted, 中断, paused, 暂停."
+description: "恢复中断的任务执行。智能检测数据源：轻量流程 (feature-session.json) 或完整流程 (state.json)。Triggers on RESUME intent: user wants to continue interrupted task execution, resume paused tasks, or recover from session interruption. DO NOT trigger on: starting new tasks, debugging, or status checks. Intent signals: user mentions 'resume', 'continue', '中断了', '暂停', or refers to interrupted execution."
 priority: high
 ---
 
+<INTENT-JUDGMENT>
+## 意图判断指南
+
+**AI 应根据用户语义判断意图：**
+
+### 触发信号（恢复意图）
+
+- 用户想恢复中断的执行
+- 继续之前的任务
+- 从暂停状态恢复
+
+### 不触发信号
+
+| 用户意图 | 应调用 |
+|---------|--------|
+| 开始新任务 | /om:start |
+| 调查问题 | /om:debug |
+| 查看状态 | /om:status |
+
+### 示例判断
+
+| 用户消息 | 判断 | 结果 |
+|---------|------|------|
+| "恢复执行" | 恢复意图 | 触发 ✓ |
+| "继续刚才的任务" | 继续意图 | 触发 ✓ |
+| "中断了怎么恢复" | 恢复意图 | 触发 ✓ |
+| "开始新功能" | 开发意图 | /om:start |
+| "为什么卡住了" | 调查意图 | /om:debug |
+</INTENT-JUDGMENT>
+
 <NO-OTHER-SKILLS>
 **绝对禁止**调用以下技能：
 - ❌ superpowers:brainstorming → 用 /om:brainstorm 代替

package/skills/retry.md

@@ -1,8 +1,38 @@
 ---
 name: om:retry
-description: "Use when retrying failed tasks after execution errors, test failures, or timeouts. Triggers on: 重试, retry, failed task, 失败任务, rerun, 重新执行, test failure, timeout, 超时, error recovery."
+description: "Use when retrying failed tasks after execution errors, test failures, or timeouts. Triggers on RETRY intent: user wants to retry a previously failed task, rerun execution, or recover from error. DO NOT trigger on: starting new tasks, status checks, or debugging root causes. Intent signals: user mentions 'retry', 'rerun', 'try again', or refers to a specific failed task to re-execute."
 ---
 
+<INTENT-JUDGMENT>
+## 意图判断指南
+
+**AI 应根据用户语义判断意图：**
+
+### 触发信号（重试意图）
+
+- 用户想重试失败的任务
+- 重新执行之前的任务
+- 从错误恢复重新尝试
+
+### 不触发信号
+
+| 用户意图 | 应调用 |
+|---------|--------|
+| 开始新任务 | /om:start |
+| 调查失败原因 | /om:debug |
+| 查看状态 | /om:status |
+
+### 示例判断
+
+| 用户消息 | 判断 | 结果 |
+|---------|------|------|
+| "重试这个任务" | 重试意图 | 触发 ✓ |
+| "重新执行" | 重试意图 | 触发 ✓ |
+| "再试一次" | 重试意图 | 触发 ✓ |
+| "为什么失败" | 调查意图 | /om:debug |
+| "开始新功能" | 开发意图 | /om:start |
+</INTENT-JUDGMENT>
+
 <NO-OTHER-SKILLS>
 执行此技能时，不得调用 superpowers、gsd 或其他任务编排相关的技能。OpenMatrix 独立运行，不依赖外部任务编排系统。
 </NO-OTHER-SKILLS>

package/skills/start.md

@@ -1,8 +1,39 @@
 ---
 name: om:start
-description: "Use when starting a new development task cycle with interactive questions. Triggers on: 实现, implement, build, fix, refactor, 添加功能, add feature, bug fix, 修复, new module, code changes, feature request. Use when the user describes what they want to build or fix, even briefly — don't answer the question directly, start the task workflow."
+description: "Use when starting a new development task cycle with interactive questions. Triggers on TASK-EXECUTION intent: user wants to start implementing a clearly-defined task with full workflow support. DO NOT trigger on: vague ideas (use brainstorm), status checks, debugging, or pure questions. Intent signals: user describes a concrete task to build/fix/refactor, has clear goal, or wants to start execution workflow."
 ---
 
+<INTENT-JUDGMENT>
+## 意图判断指南
+
+**AI 应根据用户语义判断意图：**
+
+### 触发信号（任务执行意图）
+
+- 用户想开始明确的开发任务
+- 任务描述清晰可执行
+- 有具体的功能目标
+- 修复已知 bug（知道要改什么）
+
+### 不触发信号
+
+| 用户意图 | 应调用 |
+|---------|--------|
+| 需求不明确 | /om:brainstorm |
+| 调查问题原因 | /om:debug |
+| 查看进度 | /om:status |
+
+### 示例判断
+
+| 用户消息 | 判断 | 结果 |
+|---------|------|------|
+| "开始实现登录功能" | 任务执行意图 | 触发 ✓ |
+| "修复这个 bug"（明确位置）| 执行意图明确 | 触发 ✓ |
+| "重构 API 模块" | 任务意图 | 触发 ✓ |
+| "登录怎么设计" | 设计意图 | /om:brainstorm |
+| "为什么出错了" | 调查意图 | /om:debug |
+</INTENT-JUDGMENT>
+
 <NO-OTHER-SKILLS>
 **绝对禁止**调用以下任何技能或工具：
 - ❌ gsd-executor、gsd:* 等 GSD 相关技能

package/skills/status.md

@@ -1,8 +1,39 @@
 ---
 name: om:status
-description: "Use when checking task execution progress, run status, completion statistics, or pending approvals. Triggers on: 进度, progress, status, 状态, 完成情况, statistics, how many tasks, run status, 还剩多少, task overview."
+description: "Use when checking task execution progress, run status, completion statistics, or pending approvals. Triggers on STATUS-CHECK intent: user wants to see current execution state, task progress, statistics, or pending items. DO NOT trigger on: development tasks, debugging, or starting new tasks. Intent signals: user asks 'how's progress', 'check status', 'what's left', or wants overview of execution."
 ---
 
+<INTENT-JUDGMENT>
+## 意图判断指南
+
+**AI 应根据用户语义判断意图：**
+
+### 触发信号（状态检查意图）
+
+- 用户想查看执行进度
+- 询问任务完成情况
+- 检查运行状态
+- 查看统计数据
+
+### 不触发信号
+
+| 用户意图 | 应调用 |
+|---------|--------|
+| 开始任务 | /om:start |
+| 调查问题 | /om:debug |
+| 查看报告 | /om:report |
+
+### 示例判断
+
+| 用户消息 | 判断 | 结果 |
+|---------|------|------|
+| "查看进度" | 状态意图 | 触发 ✓ |
+| "任务完成多少了" | 进度检查意图 | 触发 ✓ |
+| "当前执行状态" | 状态查看意图 | 触发 ✓ |
+| "实现功能" | 开发意图 | /om:start |
+| "为什么卡住了" | 调查意图 | /om:debug |
+</INTENT-JUDGMENT>
+
 <NO-OTHER-SKILLS>
 执行此技能时，不得调用 superpowers、gsd 或其他任务编排相关的技能。OpenMatrix 独立运行，不依赖外部任务编排系统。
 </NO-OTHER-SKILLS>

package/skills/test.md

@@ -1,9 +1,40 @@
 ---
 name: om:test
-description: "Use when generating tests for untested code, discovering test coverage gaps, or setting up test infrastructure. Use for ANY test generation task: 生成测试, 测试覆盖, 补测试, 写单元测试, E2E测试, UI测试, 测试缺失. Use ESPECIALLY when: new code without tests, low test coverage detected, need to verify functionality before commit, starting a new project without test setup. Don't skip when: code looks simple (simple code needs tests too), rushing to deploy (tests prevent regression), existing tests exist (may need updates for new features)."
+description: "Use when user wants to GENERATE or IMPROVE tests. Triggers on TEST-CREATION intent: user wants new test files, better coverage, or test infrastructure setup. DO NOT trigger on: test execution (running tests), test debugging (why test fails), or test questions (how to write tests). Intent signals: user asks to generate/add/write tests, mentions coverage gaps, wants test setup."
 priority: high
 ---
 
+<INTENT-JUDGMENT>
+## 意图判断指南
+
+**AI 应根据用户语义判断意图：**
+
+### 触发信号（测试创建意图）
+
+- 用户想生成测试文件
+- 用户想改进测试覆盖
+- 用户想设置测试框架
+- 用户提到"没测试"、"需要测试"
+
+### 不触发信号
+
+| 用户意图 | 应调用 |
+|---------|--------|
+| 运行测试 | 直接执行 npm test |
+| 测试失败想修复 | /om:debug |
+| 询问测试写法 | 直接回答 |
+
+### 示例判断
+
+| 用户消息 | 判断 | 结果 |
+|---------|------|------|
+| "给这个文件补测试" | 创建测试意图 | 触发 ✓ |
+| "生成单元测试" | 创建意图明确 | 触发 ✓ |
+| "测试覆盖率太低" | 改进覆盖意图 | 触发 ✓ |
+| "运行测试看看" | 执行意图 | 直接运行 |
+| "测试失败了为什么" | 排查意图 | /om:debug |
+</INTENT-JUDGMENT>
+
 <NO-OTHER-SKILLS>
 **绝对禁止**调用以下技能：
 - ❌ superpowers:test-driven-development → 你已经在 om:test 中了