@devflow-tools/claude-code-plugin 0.8.10 → 0.9.0

package/README.md

@@ -1,24 +1,53 @@
 # @devflow-tools/claude-code-plugin
 
-DevFlow plugin for Claude Code — provides developer intelligence platform inside Claude Code.
-
-> **Note:** This package is not published to npm directly. It's bundled within `@devflow-tools/cli` and deployed to `~/.claude/plugins/cache/`.
-
-## What You Get
-
-- **Context Engine** — Search code by natural language. Use `/devflow:context`
-- **Graph Engine** — View architecture layers and dependency graphs. Use `/devflow:graph`
-- **Memory Engine** — Access project conventions and past decisions. Use `/devflow:memory`
-- **Knowledge Engine** — Search official docs for React, Vue, NestJS, and more. Use `/devflow:knowledge`
-- **Workflow Engine** — Run standardized dev workflows. Use `/devflow:workflow`
-- **Domain Plugins** — Specialized expertise for 14 frameworks via `/devflow:react`, `/devflow:vue`, etc.
-
-## Installation
+DevFlow Claude Code 插件 — 在 Claude Code 中提供开发者智能平台的完整能力。
+
+> **Note:** 此包不直接发布到 npm。由 `@devflow-tools/cli` 打包并部署到 `~/.claude/plugins/cache/`。
+
+## 能力
+
+安装后，在 Claude Code 中可以使用以下 slash commands 和 MCP tools：
+
+### Slash Commands
+
+| 命令 | 功能 |
+|------|------|
+| `/devflow:context` | 代码上下文检索 — 自然语言查询最相关代码 |
+| `/devflow:graph` | 依赖图谱 — 查看架构分层和文件依赖 |
+| `/devflow:memory` | 项目记忆 — 约定、技术栈、历史决策 |
+| `/devflow:knowledge` | 知识库搜索 — 官方文档查询 |
+| `/devflow:workflow` | 工作流 — 标准化开发流程 |
+| `/devflow:react` | React 专家能力 — Bug 诊断、组件重构、性能审计 |
+| `/devflow:vue` | Vue 3 专家能力 |
+| `/devflow:nest` | NestJS 专家能力 |
+| `/devflow:nextjs` | Next.js 专家能力 |
+| `/devflow:electron` | Electron 专家能力 |
+| `/devflow:graphql` | GraphQL 专家能力 |
+| `/devflow:docker` | Docker 专家能力 |
+| `/devflow:css` | CSS 专家能力 |
+| `/devflow:tailwind` | Tailwind 专家能力 |
+| `/devflow:animation` | Animation 专家能力 |
+| `/devflow:performance` | Web 性能专家能力 |
+| `/devflow:taro` | Taro 跨端专家能力 |
+| `/devflow:ui-layout` | UI Layout 专家能力 |
+| `/devflow:git` | Git 专家能力 |
+
+### Hooks
+
+插件提供以下 Claude Code hooks：
+
+| Hook | 触发时机 | 功能 |
+|------|---------|------|
+| `PreToolUse` | Tool 调用前 | 记录 telemetry 事件、保存 execution context |
+| `PostToolUse` | Tool 调用后 | 回填 tool output、发送 memory event（exitCode、stderr） |
+| `Stop` | Claude Code 停止 | 完成 skill execution 记录、聚合真实统计数据 |
+
+## 安装
 
 ```bash
-# The plugin is installed automatically via devflow init
 npm install -g @devflow-tools/cli
 devflow init
+# 插件自动部署到 ~/.claude/plugins/
 ```
 
 ## License

package/dist/hooks/post-tool-use.js

@@ -1,114 +1,2 @@
-#!/usr/bin/env node
-/**
- * DevFlow PostToolUse Hook
- * 1. Records memory event with heuristic error detection
- * 2. Backfills tool_call_events output (from PreToolUse's last-event-id)
- */
-
-const http = require('http');
-const fs = require('fs');
-const path = require('path');
-const os = require('os');
-
-const SERVER_URL = process.env.DEVFLOW_SERVER_URL || 'http://127.0.0.1:13337';
-const API_KEY = process.env.DEVFLOW_API_KEY || 'dev-key-000000';
-const PROJECT_ROOT = process.env.CLAUDE_PROJECT_DIR || process.cwd();
-const SESSION_ID = `session:${PROJECT_ROOT.split('/').pop() || 'default'}`;
-
-function postJSON(urlPath, data) {
-  return new Promise((resolve, reject) => {
-    const body = JSON.stringify(data);
-    const url = new URL(urlPath, SERVER_URL);
-    const options = {
-      hostname: url.hostname,
-      port: url.port,
-      path: url.pathname,
-      method: 'POST',
-      headers: {
-        'Content-Type': 'application/json',
-        'X-API-Key': API_KEY,
-        'Content-Length': Buffer.byteLength(body),
-      },
-      timeout: 5000,
-    };
-    const req = http.request(options, (res) => {
-      let chunks = [];
-      res.on('data', (c) => chunks.push(c));
-      res.on('end', () => resolve(Buffer.concat(chunks).toString()));
-    });
-    req.on('error', () => reject(new Error('request failed')));
-    req.write(body);
-    req.end();
-  });
-}
-
-async function main() {
-  const inputJson = process.argv[2];
-  if (!inputJson) {
-    process.exit(0);
-  }
-
-  let data;
-  try {
-    data = JSON.parse(inputJson);
-  } catch {
-    process.exit(0);
-  }
-
-  const toolName = data.tool_name || '';
-  const toolInput = data.tool_input || {};
-  const toolResponse = data.tool_response || {};
-
-  // Extract command from Bash tool input
-  const command = typeof toolInput === 'object' && toolInput.command
-    ? toolInput.command
-    : '';
-
-  // Heuristic error detection from tool_response
-  const responseStr = typeof toolResponse === 'string'
-    ? toolResponse
-    : JSON.stringify(toolResponse);
-  const hasErrorSignal = /error:|traceback|fatal|killed|command not found|ENOENT/i.test(responseStr);
-  const exitCode = hasErrorSignal ? 1 : 0;
-  const stderr = hasErrorSignal ? responseStr.slice(0, 500) : null;
-
-  // 1. Record memory event
-  const evt = {
-    id: `evt_${Date.now()}_${Math.random().toString(36).slice(2, 7)}`,
-    sessionId: SESSION_ID,
-    tool: toolName,
-    command: command || toolName,
-    exitCode,
-    stderr,
-    durationMs: 0,
-    createdAt: Date.now(),
-  };
-
-  try {
-    await postJSON(`/api/memory/events?rootPath=${encodeURIComponent(PROJECT_ROOT)}`, evt);
-  } catch {}
-
-  // 2. Backfill tool_call_events output
-  try {
-    const devflowDir = path.join(os.homedir(), '.devflow');
-    const lastEventFile = path.join(devflowDir, 'last-event-id');
-    const execFile = path.join(devflowDir, 'current-execution-id');
-
-    if (fs.existsSync(lastEventFile) && fs.existsSync(execFile)) {
-      const eventId = fs.readFileSync(lastEventFile, 'utf-8').trim();
-      const execId = fs.readFileSync(execFile, 'utf-8').trim();
-
-      const output = typeof toolResponse === 'string'
-        ? { text: toolResponse.slice(0, 1000) }
-        : toolResponse;
-
-      await postJSON('/api/telemetry/tool-call/output', {
-        eventId,
-        executionId: execId,
-        output,
-      });
-    }
-  } catch {}
-}
-
-main().catch(() => process.exit(0));
+import {request}from'http';import {existsSync,readFileSync,writeFileSync}from'fs';import {join}from'path';import {homedir}from'os';var R=process.env.DEVFLOW_SERVER_URL||"http://127.0.0.1:13337",O=process.env.DEVFLOW_API_KEY||"dev-key-000000",p=process.env.CLAUDE_PROJECT_DIR||process.cwd(),u=join(homedir(),".devflow","session-id"),d=join(homedir(),".devflow","event-map.json");function m(r,n){return new Promise(e=>{let o=JSON.stringify(n),t=new URL(r,R),s=request({hostname:t.hostname,port:t.port,path:t.pathname,method:"POST",headers:{"Content-Type":"application/json","X-API-Key":O,"Content-Length":Buffer.byteLength(o)},timeout:5e3},()=>e());s.on("error",()=>e()),s.write(o),s.end();})}function P(){try{if(existsSync(u))return readFileSync(u,"utf-8").trim()}catch{}return null}function L(r){try{if(!existsSync(d))return null;let n=JSON.parse(readFileSync(d,"utf-8")),e=n.findIndex(t=>t.toolName===r);if(e===-1)return null;let o=n[e];return n.splice(e,1),writeFileSync(d,JSON.stringify(n)),{eventId:o.eventId,timestamp:o.timestamp}}catch{return null}}async function N(r){if(!r)return;let n;try{n=JSON.parse(r);}catch{return}let e=n.tool_name||"",o=n.tool_input||{},t=n.tool_response||{},s=P(),c=L(e);if(c&&s){let i=typeof t=="string"?t:JSON.stringify(t),I=i.length>5e3?{_truncated:true,_originalSize:i.length,text:i.slice(0,5e3)}:t,S=Date.now()-c.timestamp;m("/api/telemetry/tool-call/output",{eventId:c.eventId,sessionId:s,output:I,duration:S,completedAt:Date.now()}).catch(()=>{});}let v=typeof o=="object"&&o.command?o.command:"",l=typeof t=="object"&&t!==null?t:null,a=l?.exitCode,_=l?.stderr;if(e==="Bash"&&a!==void 0&&a!==0){let i={id:`evt:${Date.now()}:${Math.random().toString(36).slice(2,7)}`,sessionId:s??`session:${p.split("/").pop()||"default"}`,tool:e,command:v||e,exitCode:a,stderr:_??null,durationMs:0,createdAt:Date.now()};m(`/api/memory/events?rootPath=${encodeURIComponent(p)}`,i).catch(()=>{});}}(process.argv[1]?.endsWith("post-tool-use")||process.argv[1]?.endsWith("post-tool-use.js"))&&N(process.argv[2]||"").catch(()=>process.exit(0));
+export{N as handlePostToolUse};

package/dist/hooks/pre-tool-use.js

@@ -1,2 +1,2 @@
-import {existsSync,readFileSync,mkdirSync,writeFileSync,unlinkSync,readdirSync}from'fs';import {request}from'http';import {join}from'path';import {homedir}from'os';var k=1e4,M=300*1e3;function g(r,t,e){return new Promise(n=>{try{let o=new URL(r),i=request({hostname:o.hostname,port:o.port||80,path:o.pathname+o.search,method:"POST",headers:{"Content-Type":"application/json","X-API-Key":e,"Content-Length":Buffer.byteLength(t)},timeout:5e3},s=>{s.resume(),n(s.statusCode!=null&&s.statusCode>=200&&s.statusCode<300);});i.on("error",()=>n(!1)),i.on("timeout",()=>{i.destroy(),n(!1);}),i.write(t),i.end();}catch{n(false);}})}var l=class{constructor(t){this.retryScheduled=false;this.apiUrl=t?.apiUrl??process.env.DEVFLOW_API_URL??"http://127.0.0.1:13337",this.apiKey=t?.apiKey??process.env.DEVFLOW_API_KEY??"dev-key-000000",this.cacheDir=t?.cacheDir??join(homedir(),".devflow","telemetry-cache");}async sendEvent(t){let e=this.truncateInput(t.input),n={...t,input:e},o=`${this.apiUrl}/api/telemetry/tool-call`;await g(o,JSON.stringify(n),this.apiKey)||this.cacheEvent({type:"tool_call",payload:n});}async sendExecutionStart(t,e,n){let o={executionId:t,skillName:e,startedAt:n},i=`${this.apiUrl}/api/telemetry/skill-execution/start`;await g(i,JSON.stringify(o),this.apiKey)||this.cacheEvent({type:"execution_start",payload:o});}async flushCache(){if(!existsSync(this.cacheDir))return;let t=readdirSync(this.cacheDir).filter(e=>e.endsWith(".json"));for(let e of t)try{let n=join(this.cacheDir,e),o=readFileSync(n,"utf8"),i=JSON.parse(o),s=i.type==="execution_start"?"/api/telemetry/skill-execution/start":"/api/telemetry/tool-call",a=`${this.apiUrl}${s}`;await g(a,JSON.stringify(i.payload),this.apiKey)&&unlinkSync(n);}catch{}}cacheEvent(t){try{existsSync(this.cacheDir)||mkdirSync(this.cacheDir,{recursive:!0});let e=join(this.cacheDir,`${Date.now()}_${Math.random().toString(36).slice(2,11)}.json`);writeFileSync(e,JSON.stringify(t,null,2)),this.scheduleRetry();}catch{}}scheduleRetry(){this.retryScheduled||(this.retryScheduled=true,setTimeout(()=>{this.retryScheduled=false,this.flushCache().catch(()=>{});},M).unref());}truncateInput(t){let e=JSON.stringify(t);return e.length<=k?t:{_truncated:true,_original_size:e.length,_preview:e.substring(0,k)+"..."}}};var $=1800*1e3;function J(){return process.env.DEVFLOW_STATE_DIR??join(homedir(),".devflow")}var h=class{constructor(t){this.stateDir=J();}detect(){let t=join(this.stateDir,"current-skill.json");if(!existsSync(t))return null;try{let e=readFileSync(t,"utf8"),n=JSON.parse(e);return !n.registeredAt||Date.now()-n.registeredAt>$?(this.endExecution(),null):n}catch{return null}}startExecution(t,e){existsSync(this.stateDir)||mkdirSync(this.stateDir,{recursive:true});let n=this.getCurrentExecutionId();if(n)return n;let o=`exec_${Date.now()}_${Math.random().toString(36).slice(2,11)}`;return writeFileSync(join(this.stateDir,"current-execution-id"),o),writeFileSync(join(this.stateDir,"current-skill.json"),JSON.stringify({name:t,required_mcp_tools:e?.required_mcp_tools??[],registeredAt:Date.now()})),o}getCurrentExecutionId(){let t=join(this.stateDir,"current-execution-id");if(!existsSync(t))return null;try{return readFileSync(t,"utf8").trim()}catch{return null}}endExecution(){let t=join(this.stateDir,"current-skill.json"),e=join(this.stateDir,"current-execution-id");try{existsSync(t)&&unlinkSync(t);}catch{}try{existsSync(e)&&unlinkSync(e);}catch{}}};function p(){return {hookSpecificOutput:{hookEventName:"PreToolUse",permissionDecision:"allow"}}}function j(r){return {hookSpecificOutput:{hookEventName:"PreToolUse",permissionDecision:"deny",permissionDecisionReason:r}}}async function A(r){try{if(!r||r.trim()==="")return p();let t=JSON.parse(r),e=new h;if(t.tool_name==="Skill"){let d=t.tool_input?.skill;if(d&&d.startsWith("devflow:")){let I=e.startExecution(d);new l().sendExecutionStart(I,d,Date.now()).catch(()=>{});}}let n=e.detect(),o=t.tool_name.startsWith("mcp__devflow__"),i=o?t.tool_name.replace("mcp__devflow__",""):void 0,s=n?.required_mcp_tools??[],a=s.length>0,y=a&&!o,m;y?m=j(`Skill "${n?.name||"unknown"}" requires MCP tools. Use one of: ${s.join(", ")}`):m=p();let D=e.getCurrentExecutionId()??`exec_orphan_${Date.now()}`,w=new l,E={eventId:`evt_${Date.now()}_${Math.random().toString(36).slice(2,11)}`,executionId:D,timestamp:Date.now(),toolName:t.tool_name,toolType:o?"mcp":t.tool_name==="Agent"?"subagent":"direct",isMcpTool:o,mcpToolName:i,mcpEnforced:o&&a,mcpFallback:!o&&a,input:t.tool_input,tokensUsed:0,duration:0,blocked:!1};return w.sendEvent(E).catch(()=>{}),m}catch{return p()}}if(process.argv[1]?.endsWith("pre-tool-use")||process.argv[1]?.endsWith("pre-tool-use.js")){let r="";process.stdin.setEncoding("utf8"),process.stdin.on("data",t=>{r+=t;}),process.stdin.on("end",async()=>{let t=await A(r);console.log(JSON.stringify(t));}),process.stdin.on("error",()=>{console.log(JSON.stringify(p())),process.exit(0);}),setTimeout(()=>{console.log(JSON.stringify(p())),process.exit(0);},4e3).unref();}
-export{A as handlePreToolUse};
+import {existsSync,readdirSync,readFileSync,unlinkSync,mkdirSync,writeFileSync}from'fs';import {join}from'path';import {homedir}from'os';import {request}from'http';var I=1e4,q=3e4,W=300*1e3;function m(o,t,e){return new Promise(n=>{try{let r=new URL(o),i=request({hostname:r.hostname,port:r.port||80,path:r.pathname+r.search,method:"POST",headers:{"Content-Type":"application/json","X-API-Key":e,"Content-Length":Buffer.byteLength(t)},timeout:5e3},s=>{let c=[];s.on("data",a=>c.push(a)),s.on("end",()=>{let a=Buffer.concat(c).toString();n(s.statusCode!=null&&s.statusCode>=200&&s.statusCode<300?a:null);});});i.on("error",()=>n(null)),i.on("timeout",()=>{i.destroy(),n(null);}),i.write(t),i.end();}catch{n(null);}})}var f=class{constructor(t){this.retryScheduled=false;this.retryAttempts=0;this.apiUrl=t?.apiUrl??process.env.DEVFLOW_API_URL??"http://127.0.0.1:13337",this.apiKey=t?.apiKey??process.env.DEVFLOW_API_KEY??"dev-key-000000",this.cacheDir=t?.cacheDir??join(homedir(),".devflow","telemetry-cache");}async sendEvent(t){let e=this.truncateInput(t.input),n={...t,input:e},r=`${this.apiUrl}/api/telemetry/tool-call`,i=await m(r,JSON.stringify(n),this.apiKey);if(i)try{let s=JSON.parse(i),c=s.data?.eventId??s.eventId??null;if(c)return c}catch{}return this.cacheEvent({type:"tool_call",payload:n}),null}async sendExecutionStart(t,e,n,r){let i={executionId:t,sessionId:e,skillName:n,startedAt:r},s=`${this.apiUrl}/api/telemetry/skill-execution/start`;await m(s,JSON.stringify(i),this.apiKey)||this.cacheEvent({type:"execution_start",payload:i});}async sendSessionStart(t,e,n){let r={id:t,projectRoot:e,startedAt:n,label:e.split("/").pop()??"unknown"},i=`${this.apiUrl}/api/telemetry/sessions`;return await m(i,JSON.stringify(r),this.apiKey)!==null}async flushCache(){if(!existsSync(this.cacheDir))return;let t=readdirSync(this.cacheDir).filter(e=>e.endsWith(".json"));for(let e of t)try{let n=join(this.cacheDir,e),r=readFileSync(n,"utf8"),i=JSON.parse(r),s=i.type==="execution_start"?"/api/telemetry/skill-execution/start":"/api/telemetry/tool-call",c=`${this.apiUrl}${s}`;await m(c,JSON.stringify(i.payload),this.apiKey)&&unlinkSync(n);}catch{}}cacheEvent(t){try{existsSync(this.cacheDir)||mkdirSync(this.cacheDir,{recursive:!0});let e=join(this.cacheDir,`${Date.now()}_${Math.random().toString(36).slice(2,11)}.json`);writeFileSync(e,JSON.stringify(t,null,2)),this.scheduleRetry();}catch{}}scheduleRetry(){if(this.retryScheduled)return;this.retryScheduled=true;let t=Math.min(q*Math.pow(2,this.retryAttempts),W);setTimeout(()=>{this.retryScheduled=false,this.flushCache().then(()=>{this.retryAttempts=0;}).catch(()=>{this.retryAttempts++;});},t).unref();}truncateInput(t){let e=JSON.stringify(t);return e.length<=I?t:{_truncated:true,_original_size:e.length,_preview:e.substring(0,I)+"..."}}};var V=1800*1e3;function X(){return process.env.DEVFLOW_STATE_DIR??join(homedir(),".devflow")}var g=class{constructor(t){this.stateDir=X();}detect(){let t=join(this.stateDir,"current-skill.json");if(!existsSync(t))return null;try{let e=readFileSync(t,"utf8"),n=JSON.parse(e);return !n.registeredAt||Date.now()-n.registeredAt>V?(this.endExecution(),null):n}catch{return null}}startExecution(t,e){existsSync(this.stateDir)||mkdirSync(this.stateDir,{recursive:true});let n=this.getCurrentExecutionId();if(n)return n;let r=`exec_${Date.now()}_${Math.random().toString(36).slice(2,11)}`;return writeFileSync(join(this.stateDir,"current-execution-id"),r),writeFileSync(join(this.stateDir,"current-skill.json"),JSON.stringify({name:t,required_mcp_tools:e?.required_mcp_tools??[],registeredAt:Date.now()})),r}getCurrentExecutionId(){let t=join(this.stateDir,"current-execution-id");if(!existsSync(t))return null;try{return readFileSync(t,"utf8").trim()}catch{return null}}endExecution(){let t=join(this.stateDir,"current-skill.json"),e=join(this.stateDir,"current-execution-id");try{existsSync(t)&&unlinkSync(t);}catch{}try{existsSync(e)&&unlinkSync(e);}catch{}}};function d(){return {hookSpecificOutput:{hookEventName:"PreToolUse",permissionDecision:"allow"}}}function G(o){return {hookSpecificOutput:{hookEventName:"PreToolUse",permissionDecision:"deny",permissionDecisionReason:o}}}var h=join(homedir(),".devflow");function A(){existsSync(h)||mkdirSync(h,{recursive:true});}function Z(){try{let o=join(h,"session-id");if(existsSync(o))return readFileSync(o,"utf-8").trim()}catch{}return null}function Q(o){A(),writeFileSync(join(h,"session-id"),o);}function tt(o,t,e){A();let n=join(h,"event-map.json"),r=[];try{existsSync(n)&&(r=JSON.parse(readFileSync(n,"utf-8")));}catch{}r.push({toolName:o,eventId:t,timestamp:e}),writeFileSync(n,JSON.stringify(r));}function et(o){switch(o){case "Read":case "Glob":return "file_read";case "Write":case "Edit":case "NotebookEdit":return "file_write";case "Bash":return "bash_command";case "Agent":return "subagent";case "Skill":return "skill_invoke";default:return "tool_use"}}async function nt(o){try{if(!o||o.trim()==="")return d();let t=JSON.parse(o),e=process.env.CLAUDE_PROJECT_DIR||process.cwd(),n=e.split("/").pop()||"unknown",r=new f,i=Z();i||(i=`session:${n}:${Date.now().toString(36)}`,Q(i),r.sendSessionStart(i,e,Date.now()).catch(()=>{}));let s=new g;if(t.tool_name==="Skill"){let l=t.tool_input?.skill;if(l&&l.startsWith("devflow:")){let C=s.startExecution(l);r.sendExecutionStart(C,i,l,Date.now()).catch(()=>{});}}let c=s.detect(),a=t.tool_name.startsWith("mcp__devflow__"),M=a?t.tool_name.replace("mcp__devflow__",""):void 0,w=c?.required_mcp_tools??[],_=w.length>0,D=_&&!a,S;D?S=G(`Skill "${c?.name||"unknown"}" requires MCP tools. Use one of: ${w.join(", ")}`):S=d();let R=s.getCurrentExecutionId()??i,E={eventId:`evt_${Date.now()}_${Math.random().toString(36).slice(2,11)}`,executionId:R,sessionId:i,timestamp:Date.now(),toolName:t.tool_name,toolType:a?"mcp":t.tool_name==="Agent"?"subagent":"direct",isMcpTool:a,mcpToolName:M,mcpEnforced:a&&_,mcpFallback:!a&&_,kind:et(t.tool_name),input:t.tool_input,duration:0,blocked:D};return r.sendEvent(E).then(l=>{l&&tt(t.tool_name,l,E.timestamp);}).catch(()=>{}),S}catch{return d()}}if(process.argv[1]?.endsWith("pre-tool-use")||process.argv[1]?.endsWith("pre-tool-use.js")){let o="";process.stdin.setEncoding("utf8"),process.stdin.on("data",t=>{o+=t;}),process.stdin.on("end",async()=>{let t=await nt(o);console.log(JSON.stringify(t));}),process.stdin.on("error",()=>{console.log(JSON.stringify(d())),process.exit(0);}),setTimeout(()=>{console.log(JSON.stringify(d())),process.exit(0);},4e3).unref();}
+export{nt as handlePreToolUse};

package/dist/hooks/stop.js

@@ -1,118 +1,2 @@
-#!/usr/bin/env node
-/**
- * DevFlow Stop Hook
- * 1. Completes pending skill execution with REAL aggregated data
- * 2. Triggers memory extraction for the session
- * 3. Outputs violation summary
- */
-
-const http = require('http');
-const fs = require('fs');
-const path = require('path');
-const os = require('os');
-
-const SERVER_URL = process.env.DEVFLOW_SERVER_URL || 'http://127.0.0.1:13337';
-const API_KEY = process.env.DEVFLOW_API_KEY || 'dev-key-000000';
-const PROJECT_ROOT = process.env.CLAUDE_PROJECT_DIR || process.cwd();
-
-function getJSON(urlPath) {
-  return new Promise((resolve, reject) => {
-    const url = new URL(urlPath, SERVER_URL);
-    http.get(url.toString(), { headers: { 'X-API-Key': API_KEY }, timeout: 5000 }, (res) => {
-      let chunks = [];
-      res.on('data', (c) => chunks.push(c));
-      res.on('end', () => {
-        try { resolve(JSON.parse(Buffer.concat(chunks).toString())); }
-        catch { resolve(null); }
-      });
-    }).on('error', () => resolve(null));
-  });
-}
-
-function postJSON(urlPath, data) {
-  return new Promise((resolve) => {
-    const body = JSON.stringify(data);
-    const url = new URL(urlPath, SERVER_URL);
-    const req = http.request(url.toString(), {
-      method: 'POST',
-      headers: { 'Content-Type': 'application/json', 'X-API-Key': API_KEY, 'Content-Length': Buffer.byteLength(body) },
-      timeout: 5000,
-    }, (res) => { resolve(null); });
-    req.on('error', () => resolve(null));
-    req.write(body);
-    req.end();
-  });
-}
-
-async function main() {
-  const devflowDir = path.join(os.homedir(), '.devflow');
-  const execFile = path.join(devflowDir, 'current-execution-id');
-  const skillFile = path.join(devflowDir, 'current-skill.json');
-
-  // ── 1. Complete pending skill execution with REAL counts ──
-  if (fs.existsSync(execFile)) {
-    const execId = fs.readFileSync(execFile, 'utf-8').trim();
-    if (execId) {
-      try {
-        // Query real event data from the API
-        const result = await getJSON(`/api/telemetry/executions/${execId}/events`);
-        const events = result?.data || [];
-
-        const totalToolCalls = events.length;
-        const mcpToolCalls = events.filter(e => e.isMcpTool).length;
-        const directToolCalls = totalToolCalls - mcpToolCalls;
-        const subagentCount = events.filter(e => e.toolType === 'subagent').length;
-        const totalTokens = events.reduce((sum, e) => sum + (e.tokensUsed || 0), 0);
-
-        const timestamps = events.map(e => e.timestamp).filter(Boolean);
-        const totalDuration = timestamps.length >= 2
-          ? Math.max(...timestamps) - Math.min(...timestamps)
-          : 0;
-
-        const mcpRate = totalToolCalls > 0
-          ? Math.round((mcpToolCalls / totalToolCalls) * 10000) / 100
-          : 0;
-
-        const nowMs = Date.now();
-
-        await postJSON('/api/telemetry/skill-execution/complete', {
-          executionId: execId,
-          finishedAt: nowMs,
-          status: 'completed',
-          summary: {
-            totalToolCalls,
-            mcpToolCalls,
-            directToolCalls,
-            subagentCount,
-            totalTokens,
-            totalDuration,
-            mcpComplianceRate: mcpRate,
-          },
-        });
-      } catch (e) {
-        console.error('[devflow] Stop hook error:', e.message);
-      }
-    }
-    try { fs.unlinkSync(execFile); } catch {}
-    try { fs.unlinkSync(skillFile); } catch {}
-  }
-
-  // ── 2. Trigger memory extraction ──
-  try {
-    await postJSON(`/api/memory/process?rootPath=${encodeURIComponent(PROJECT_ROOT)}`, {});
-  } catch {}
-
-  // ── 3. Violation summary ──
-  try {
-    const sessionId = require('crypto').createHash('sha256').update(PROJECT_ROOT).digest('hex').slice(0, 16);
-    const receiptFile = path.join(os.tmpdir(), '.devflow-receipts', `${sessionId}.json`);
-    if (fs.existsSync(receiptFile)) {
-      const receipt = JSON.parse(fs.readFileSync(receiptFile, 'utf-8'));
-      const bypass = receipt.bypassCount || 0;
-      console.error(`会话违规统计: ${bypass}/3`);
-      try { fs.unlinkSync(receiptFile); } catch {}
-    }
-  } catch {}
-}
-
-main().catch(() => process.exit(0));
+import {get,request}from'http';import {existsSync,readFileSync,unlinkSync}from'fs';import {join}from'path';import {homedir}from'os';var I=process.env.DEVFLOW_SERVER_URL||"http://127.0.0.1:13337",x=process.env.DEVFLOW_API_KEY||"dev-key-000000",D=process.env.CLAUDE_PROJECT_DIR||process.cwd();function b(t){return new Promise(r=>{let e=new URL(t,I);get(e.toString(),{headers:{"X-API-Key":x},timeout:5e3},n=>{let c=[];n.on("data",g=>c.push(g)),n.on("end",()=>{try{r(JSON.parse(Buffer.concat(c).toString()));}catch{r(null);}});}).on("error",()=>r(null));})}function d(t,r){return new Promise(e=>{let s=JSON.stringify(r),n=new URL(t,I),c=request({hostname:n.hostname,port:n.port,path:n.pathname,method:"POST",headers:{"Content-Type":"application/json","X-API-Key":x,"Content-Length":Buffer.byteLength(s)},timeout:5e3},()=>e());c.on("error",()=>e()),c.write(s),c.end();})}var p=join(homedir(),".devflow"),E=join(p,"session-id"),T=join(p,"current-execution-id"),M=join(p,"current-skill.json"),O=join(p,"event-map.json");function A(){for(let t of [E,T,M,O])try{existsSync(t)&&unlinkSync(t);}catch{}}async function F(){let t=(()=>{try{if(existsSync(E))return readFileSync(E,"utf-8").trim()}catch{}return null})();if(t){let e=(await b(`/api/telemetry/sessions/${t}/events`))?.data||[];e.length;e.filter(o=>o.isMcpTool).length;e.filter(o=>o.toolType==="subagent").length;let u=e.map(o=>o.timestamp).filter(Boolean);u.length>=2?Math.max(...u)-Math.min(...u):0;if(existsSync(T)){let o="";try{o=readFileSync(T,"utf-8").trim();}catch{}if(o){let l=e.filter(a=>a.executionId===o),m=l.length,h=l.filter(a=>a.isMcpTool).length,S=m-h,R=l.filter(a=>a.toolType==="subagent").length,f=l.map(a=>a.timestamp).filter(Boolean),_=f.length>=2?Math.max(...f)-Math.min(...f):0;try{await d("/api/telemetry/skill-execution/complete",{executionId:o,finishedAt:Date.now(),status:"completed",summary:{totalToolCalls:m,mcpToolCalls:h,directToolCalls:S,subagentCount:R,totalDuration:_,mcpComplianceRate:m>0?Math.round(h/m*1e4)/100:0}});}catch{}}}await d(`/api/telemetry/sessions/${t}/close`,{finishedAt:Date.now()});}d(`/api/memory/process?rootPath=${encodeURIComponent(D)}`,{}).catch(()=>{}),A();}(process.argv[1]?.endsWith("stop")||process.argv[1]?.endsWith("stop.js"))&&F().catch(()=>process.exit(0));
+export{F as handleStop};

package/dist/skills/devflow:context/SKILL.md

@@ -11,7 +11,6 @@ required_mcp_tools:
 get_project_context / search_symbol
 </HARD-GATE>
 
-
 ## MANDATORY FIRST STEP - Skill Registration
 
 > **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
@@ -31,7 +30,6 @@ When the skill execution is complete, clean up:
 rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
 ```
 
-
 # /devflow:context
 
 ## 执行

package/dist/skills/devflow:docker/SKILL.md

@@ -10,7 +10,6 @@ required_mcp_tools:
 docker_dockerize_project
 </HARD-GATE>
 
-
 ## MANDATORY FIRST STEP - Skill Registration
 
 > **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
@@ -29,7 +28,6 @@ When the skill execution is complete, clean up:
 rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
 ```
 
-
 # 🐳 DOCKER
 
 你是一个 Docker 专家。使用多阶段构建优化镜像，遵循安全和性能最佳实践。遇到问题先查官方文档。

package/dist/skills/devflow:electron/SKILL.md

@@ -10,7 +10,6 @@ required_mcp_tools:
 electron_diagnose_bug
 </HARD-GATE>
 
-
 ## MANDATORY FIRST STEP - Skill Registration
 
 > **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
@@ -29,7 +28,6 @@ When the skill execution is complete, clean up:
 rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
 ```
 
-
 # ⚛️ ELECTRON
 
 你是一个 Electron 专家。使用 contextBridge + IPC 安全通信模式。严格遵循 Electron 安全最佳实践。

package/dist/skills/devflow:graph/SKILL.md

@@ -10,7 +10,6 @@ required_mcp_tools:
 get_dependency_graph
 </HARD-GATE>
 
-
 ## MANDATORY FIRST STEP - Skill Registration
 
 > **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
@@ -29,7 +28,6 @@ When the skill execution is complete, clean up:
 rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
 ```
 
-
 # /devflow:graph
 
 ## 执行

package/dist/skills/devflow:graphql/SKILL.md

@@ -11,7 +11,6 @@ required_mcp_tools:
 graphql_new_graphql_type / graphql_diagnose_bug
 </HARD-GATE>
 
-
 ## MANDATORY FIRST STEP - Skill Registration
 
 > **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
@@ -31,7 +30,6 @@ When the skill execution is complete, clean up:
 rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
 ```
 
-
 # ◈ GRAPHQL
 
 你是一个 GraphQL 专家。使用 Schema-First 设计，Apollo Server 实现。遇到问题先查文档，不凭空猜测。

package/dist/skills/devflow:knowledge/SKILL.md

@@ -10,7 +10,6 @@ required_mcp_tools:
 get_knowledge
 </HARD-GATE>
 
-
 ## MANDATORY FIRST STEP - Skill Registration
 
 > **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
@@ -29,7 +28,6 @@ When the skill execution is complete, clean up:
 rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
 ```
 
-
 # /devflow:knowledge
 
 ## 执行

package/dist/skills/devflow:memory/SKILL.md

@@ -10,7 +10,6 @@ required_mcp_tools:
 get_memory
 </HARD-GATE>
 
-
 ## MANDATORY FIRST STEP - Skill Registration
 
 > **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
@@ -29,7 +28,6 @@ When the skill execution is complete, clean up:
 rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
 ```
 
-
 # /devflow:memory
 
 ## 执行

package/dist/skills/devflow:nest/SKILL.md

@@ -11,7 +11,6 @@ required_mcp_tools:
 nest_new_module / nest_diagnose_bug
 </HARD-GATE>
 
-
 ## MANDATORY FIRST STEP - Skill Registration
 
 > **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
@@ -31,7 +30,6 @@ When the skill execution is complete, clean up:
 rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
 ```
 
-
 # 🐱 NEST
 
 你是一个 NestJS 专家。使用装饰器模式、依赖注入和模块化架构。遵循 SOLID 原则。

package/dist/skills/devflow:nextjs/SKILL.md

@@ -11,7 +11,6 @@ required_mcp_tools:
 nextjs_new_page / nextjs_diagnose_bug
 </HARD-GATE>
 
-
 ## MANDATORY FIRST STEP - Skill Registration
 
 > **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
@@ -31,7 +30,6 @@ When the skill execution is complete, clean up:
 rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
 ```
 
-
 # ▲ NEXTJS
 
 你是一个 Next.js 14 专家。使用 App Router、React Server Components。默认服务端渲染，按需客户端交互。

package/dist/skills/devflow:performance/SKILL.md

@@ -10,7 +10,6 @@ required_mcp_tools:
 performance_audit_performance
 </HARD-GATE>
 
-
 ## MANDATORY FIRST STEP - Skill Registration
 
 > **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
@@ -29,7 +28,6 @@ When the skill execution is complete, clean up:
 rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
 ```
 
-
 # ⚡ PERFORMANCE
 
 你是一个 Web 性能专家。精通 Core Web Vitals、Bundle 优化和运行时性能。所有优化建议必须有数据依据。

package/dist/skills/devflow:react/SKILL.md

@@ -15,7 +15,6 @@ required_mcp_tools:
 react_diagnose_bug / react_new_feature / react_audit_performance / react_review_hooks / react_refactor_component / react_new_component
 </HARD-GATE>
 
-
 ## MANDATORY FIRST STEP - Skill Registration
 
 > **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
@@ -39,7 +38,6 @@ When the skill execution is complete, clean up:
 rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
 ```
 
-
 # ⚛️ REACT
 
 你是一个 React 18 专家。使用函数组件、Hooks 和 TypeScript。遇到问题先查文档再给方案，不凭空猜测 API 行为。

package/dist/skills/devflow:tailwind/SKILL.md

@@ -10,7 +10,6 @@ required_mcp_tools:
 tailwind_new_component
 </HARD-GATE>
 
-
 ## MANDATORY FIRST STEP - Skill Registration
 
 > **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
@@ -29,7 +28,6 @@ When the skill execution is complete, clean up:
 rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
 ```
 
-
 # 🌊 TAILWIND
 
 你是一个 Tailwind CSS 专家。使用 utility-first 方法构建界面。遇到问题先查 Tailwind 官方文档。

package/dist/skills/devflow:taro/SKILL.md

@@ -10,7 +10,6 @@ required_mcp_tools:
 taro_new_page
 </HARD-GATE>
 
-
 ## MANDATORY FIRST STEP - Skill Registration
 
 > **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
@@ -29,7 +28,6 @@ When the skill execution is complete, clean up:
 rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
 ```
 
-
 # 🛠️ TARO
 
 你是一个 Taro 跨端开发专家。使用 Taro 内置组件和 API，确保多端兼容。遇到小程序问题先查文档，不凭经验猜测。

package/dist/skills/devflow:vue/SKILL.md

@@ -12,7 +12,6 @@ required_mcp_tools:
 vue_diagnose_bug / vue_new_component / vue_debug_reactivity
 </HARD-GATE>
 
-
 ## MANDATORY FIRST STEP - Skill Registration
 
 > **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
@@ -33,7 +32,6 @@ When the skill execution is complete, clean up:
 rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
 ```
 
-
 # 💚 VUE
 
 你是一个 Vue 3 专家。使用 Composition API、<script setup> 和 TypeScript。遇到问题先查文档再给方案，不凭空猜测。

package/dist/skills/devflow:workflow/SKILL.md

@@ -10,7 +10,6 @@ required_mcp_tools:
 run_workflow
 </HARD-GATE>
 
-
 ## MANDATORY FIRST STEP - Skill Registration
 
 > **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
@@ -29,7 +28,6 @@ When the skill execution is complete, clean up:
 rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
 ```
 
-
 # /devflow:workflow
 
 ## 执行

package/hooks/post-tool-use

@@ -5,7 +5,8 @@ INPUT_JSON=$(cat)
 PROJECT_ROOT="${CLAUDE_PROJECT_DIR:-$(pwd)}"
 
 # ── Event Recording via Node.js (best-effort, never blocks hook) ──
-HOOK_JS="${CLAUDE_PLUGIN_ROOT:-}/dist/hooks/post-tool-use.js"
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+HOOK_JS="${CLAUDE_PLUGIN_ROOT:-$SCRIPT_DIR/..}/dist/hooks/post-tool-use.js"
 if [ -f "$HOOK_JS" ]; then
   node "$HOOK_JS" "$INPUT_JSON" 2>/dev/null || true
 fi

package/hooks/pre-tool-use

@@ -4,7 +4,8 @@ set -euo pipefail
 # DevFlow PreToolUse Hook — Node.js handler
 # Falls back to allow if Node.js handler is not available
 
-HOOK_HANDLER="${CLAUDE_PLUGIN_ROOT:-}/dist/hooks/pre-tool-use.js"
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+HOOK_HANDLER="${CLAUDE_PLUGIN_ROOT:-$SCRIPT_DIR/..}/dist/hooks/pre-tool-use.js"
 
 if [ -f "$HOOK_HANDLER" ]; then
   # Pass stdin through to Node.js handler; fallback to allow on any failure

package/hooks/stop

@@ -1,7 +1,8 @@
 #!/usr/bin/env bash
 set -euo pipefail
 
-HOOK_JS="${CLAUDE_PLUGIN_ROOT:-}/dist/hooks/stop.js"
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+HOOK_JS="${CLAUDE_PLUGIN_ROOT:-$SCRIPT_DIR/..}/dist/hooks/stop.js"
 
 if [ -f "$HOOK_JS" ]; then
   node "$HOOK_JS" 2>/dev/null || true

package/hooks.json

@@ -35,6 +35,18 @@
           }
         ]
       }
+    ],
+    "Stop": [
+      {
+        "matcher": "",
+        "hooks": [
+          {
+            "type": "command",
+            "command": "bash \"${CLAUDE_PLUGIN_ROOT}/hooks/stop\"",
+            "async": false
+          }
+        ]
+      }
     ]
   }
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@devflow-tools/claude-code-plugin",
-  "version": "0.8.10",
+  "version": "0.9.0",
   "description": "DevFlow — Developer Intelligence Platform for Claude Code",
   "type": "module",
   "scripts": {
@@ -8,7 +8,7 @@
     "deploy": "npx tsx scripts/inject-gates.ts --deploy"
   },
   "devDependencies": {
-    "@devflow-tools/mcp-server": "^0.8.10"
+    "@devflow-tools/mcp-server": "^0.9.0"
   },
   "keywords": [
     "pi-package",
@@ -28,5 +28,5 @@
     "hooks.json",
     "skills"
   ],
-  "gitHead": "ac4f93d6a52fa40b4126b514e899eee581f8f46e"
+  "gitHead": "7872bccedbacd09cc92b983f30b3702db6878125"
 }

package/skills/devflow:animation/SKILL.md

@@ -1,7 +1,6 @@
 ---
 name: devflow:animation
 description: Animation 动画专家 — CSS 动画、Framer Motion、GSAP、Lottie、性能优化
-required_mcp_tools: []
 ---
 
 # ✨ ANIMATION

package/skills/devflow:css/SKILL.md

@@ -1,7 +1,6 @@
 ---
 name: devflow:css
 description: CSS 专家 — 布局调试、响应式审查、BEM 规范、性能优化
-required_mcp_tools: []
 ---
 
 # 🎨 CSS

package/skills/devflow:docker/SKILL.md

@@ -1,27 +1,7 @@
 ---
 name: devflow:docker
 description: Docker 专家 — 镜像优化、Compose 编排、安全扫描、多阶段构建
-required_mcp_tools:
-  - docker_dockerize_project
 ---
-## MANDATORY FIRST STEP - Skill Registration
-
-> **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
-
-```bash
-mkdir -p ~/.devflow && echo '{"name":"devflow:docker","required_mcp_tools":["docker_dockerize_project"],"description":"devflow:docker skill","registeredAt":'"$(date +%s)000"'}' > ~/.devflow/current-skill.json
-```
-
-After registration, you **MUST** use ONLY these MCP tools for all work in this skill:
-- `mcp__devflow__docker_dockerize_project`
-
-**Never use direct tools** (Read, Write, Edit, Bash, Grep, etc.) for tasks that can be handled by the MCP tools above.
-
-When the skill execution is complete, clean up:
-```bash
-rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
-```
-
 
 # 🐳 DOCKER
 

package/skills/devflow:electron/SKILL.md

@@ -1,27 +1,7 @@
 ---
 name: devflow:electron
 description: Electron 专家 — IPC 通信、窗口管理、原生模块、自动更新、安全配置
-required_mcp_tools:
-  - electron_diagnose_bug
 ---
-## MANDATORY FIRST STEP - Skill Registration
-
-> **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
-
-```bash
-mkdir -p ~/.devflow && echo '{"name":"devflow:electron","required_mcp_tools":["electron_diagnose_bug"],"description":"devflow:electron skill","registeredAt":'"$(date +%s)000"'}' > ~/.devflow/current-skill.json
-```
-
-After registration, you **MUST** use ONLY these MCP tools for all work in this skill:
-- `mcp__devflow__electron_diagnose_bug`
-
-**Never use direct tools** (Read, Write, Edit, Bash, Grep, etc.) for tasks that can be handled by the MCP tools above.
-
-When the skill execution is complete, clean up:
-```bash
-rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
-```
-
 
 # ⚛️ ELECTRON
 

package/skills/devflow:git/SKILL.md

@@ -1,7 +1,6 @@
 ---
 name: devflow:git
 description: Git 专家 — 提交规范、分支审查、Changelog 生成、冲突解决
-required_mcp_tools: []
 ---
 
 # 🔀 GIT

package/skills/devflow:graphql/SKILL.md

@@ -1,29 +1,7 @@
 ---
 name: devflow:graphql
 description: GraphQL 专家 — Schema 设计、N+1 检测、Resolver 优化、安全审计
-required_mcp_tools:
-  - graphql_new_graphql_type
-  - graphql_diagnose_bug
 ---
-## MANDATORY FIRST STEP - Skill Registration
-
-> **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
-
-```bash
-mkdir -p ~/.devflow && echo '{"name":"devflow:graphql","required_mcp_tools":["graphql_new_graphql_type","graphql_diagnose_bug"],"description":"devflow:graphql skill","registeredAt":'"$(date +%s)000"'}' > ~/.devflow/current-skill.json
-```
-
-After registration, you **MUST** use ONLY these MCP tools for all work in this skill:
-- `mcp__devflow__graphql_new_graphql_type`
-- `mcp__devflow__graphql_diagnose_bug`
-
-**Never use direct tools** (Read, Write, Edit, Bash, Grep, etc.) for tasks that can be handled by the MCP tools above.
-
-When the skill execution is complete, clean up:
-```bash
-rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
-```
-
 
 # ◈ GRAPHQL
 

package/skills/devflow:nest/SKILL.md

@@ -1,29 +1,7 @@
 ---
 name: devflow:nest
 description: NestJS 专家 — 模块架构、守卫调试、微服务配置、Swagger 生成
-required_mcp_tools:
-  - nest_new_module
-  - nest_diagnose_bug
 ---
-## MANDATORY FIRST STEP - Skill Registration
-
-> **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
-
-```bash
-mkdir -p ~/.devflow && echo '{"name":"devflow:nest","required_mcp_tools":["nest_new_module","nest_diagnose_bug"],"description":"devflow:nest skill","registeredAt":'"$(date +%s)000"'}' > ~/.devflow/current-skill.json
-```
-
-After registration, you **MUST** use ONLY these MCP tools for all work in this skill:
-- `mcp__devflow__nest_new_module`
-- `mcp__devflow__nest_diagnose_bug`
-
-**Never use direct tools** (Read, Write, Edit, Bash, Grep, etc.) for tasks that can be handled by the MCP tools above.
-
-When the skill execution is complete, clean up:
-```bash
-rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
-```
-
 
 # 🐱 NEST
 

package/skills/devflow:nextjs/SKILL.md

@@ -1,29 +1,7 @@
 ---
 name: devflow:nextjs
 description: Next.js 专家 — App Router、SSR/SSG/ISR、Server Components、Middleware、部署
-required_mcp_tools:
-  - nextjs_new_page
-  - nextjs_diagnose_bug
 ---
-## MANDATORY FIRST STEP - Skill Registration
-
-> **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
-
-```bash
-mkdir -p ~/.devflow && echo '{"name":"devflow:nextjs","required_mcp_tools":["nextjs_new_page","nextjs_diagnose_bug"],"description":"devflow:nextjs skill","registeredAt":'"$(date +%s)000"'}' > ~/.devflow/current-skill.json
-```
-
-After registration, you **MUST** use ONLY these MCP tools for all work in this skill:
-- `mcp__devflow__nextjs_new_page`
-- `mcp__devflow__nextjs_diagnose_bug`
-
-**Never use direct tools** (Read, Write, Edit, Bash, Grep, etc.) for tasks that can be handled by the MCP tools above.
-
-When the skill execution is complete, clean up:
-```bash
-rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
-```
-
 
 # ▲ NEXTJS
 

package/skills/devflow:performance/SKILL.md

@@ -1,27 +1,7 @@
 ---
 name: devflow:performance
 description: Web 性能专家 — Core Web Vitals、Bundle 优化、加载策略、渲染性能
-required_mcp_tools:
-  - performance_audit_performance
 ---
-## MANDATORY FIRST STEP - Skill Registration
-
-> **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
-
-```bash
-mkdir -p ~/.devflow && echo '{"name":"devflow:performance","required_mcp_tools":["performance_audit_performance"],"description":"devflow:performance skill","registeredAt":'"$(date +%s)000"'}' > ~/.devflow/current-skill.json
-```
-
-After registration, you **MUST** use ONLY these MCP tools for all work in this skill:
-- `mcp__devflow__performance_audit_performance`
-
-**Never use direct tools** (Read, Write, Edit, Bash, Grep, etc.) for tasks that can be handled by the MCP tools above.
-
-When the skill execution is complete, clean up:
-```bash
-rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
-```
-
 
 # ⚡ PERFORMANCE
 

package/skills/devflow:react/SKILL.md

@@ -1,37 +1,7 @@
 ---
 name: devflow:react
 description: React 18 专家能力 — Bug 诊断、组件重构、性能优化、Hooks 审查
-required_mcp_tools:
-  - react_diagnose_bug
-  - react_new_feature
-  - react_audit_performance
-  - react_review_hooks
-  - react_refactor_component
-  - react_new_component
 ---
-## MANDATORY FIRST STEP - Skill Registration
-
-> **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
-
-```bash
-mkdir -p ~/.devflow && echo '{"name":"devflow:react","required_mcp_tools":["react_diagnose_bug","react_new_feature","react_audit_performance","react_review_hooks","react_refactor_component","react_new_component"],"description":"devflow:react skill","registeredAt":'"$(date +%s)000"'}' > ~/.devflow/current-skill.json
-```
-
-After registration, you **MUST** use ONLY these MCP tools for all work in this skill:
-- `mcp__devflow__react_diagnose_bug`
-- `mcp__devflow__react_new_feature`
-- `mcp__devflow__react_audit_performance`
-- `mcp__devflow__react_review_hooks`
-- `mcp__devflow__react_refactor_component`
-- `mcp__devflow__react_new_component`
-
-**Never use direct tools** (Read, Write, Edit, Bash, Grep, etc.) for tasks that can be handled by the MCP tools above.
-
-When the skill execution is complete, clean up:
-```bash
-rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
-```
-
 
 # ⚛️ REACT
 

package/skills/devflow:tailwind/SKILL.md

@@ -1,27 +1,7 @@
 ---
 name: devflow:tailwind
 description: Tailwind 专家 — 响应式设计、主题定制、Purge 检查、Utility 审查
-required_mcp_tools:
-  - tailwind_new_component
 ---
-## MANDATORY FIRST STEP - Skill Registration
-
-> **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
-
-```bash
-mkdir -p ~/.devflow && echo '{"name":"devflow:tailwind","required_mcp_tools":["tailwind_new_component"],"description":"devflow:tailwind skill","registeredAt":'"$(date +%s)000"'}' > ~/.devflow/current-skill.json
-```
-
-After registration, you **MUST** use ONLY these MCP tools for all work in this skill:
-- `mcp__devflow__tailwind_new_component`
-
-**Never use direct tools** (Read, Write, Edit, Bash, Grep, etc.) for tasks that can be handled by the MCP tools above.
-
-When the skill execution is complete, clean up:
-```bash
-rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
-```
-
 
 # 🌊 TAILWIND
 

package/skills/devflow:taro/SKILL.md

@@ -1,27 +1,7 @@
 ---
 name: devflow:taro
 description: Taro 跨端专家 — 多端适配、小程序调试、性能优化、兼容方案
-required_mcp_tools:
-  - taro_new_page
 ---
-## MANDATORY FIRST STEP - Skill Registration
-
-> **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
-
-```bash
-mkdir -p ~/.devflow && echo '{"name":"devflow:taro","required_mcp_tools":["taro_new_page"],"description":"devflow:taro skill","registeredAt":'"$(date +%s)000"'}' > ~/.devflow/current-skill.json
-```
-
-After registration, you **MUST** use ONLY these MCP tools for all work in this skill:
-- `mcp__devflow__taro_new_page`
-
-**Never use direct tools** (Read, Write, Edit, Bash, Grep, etc.) for tasks that can be handled by the MCP tools above.
-
-When the skill execution is complete, clean up:
-```bash
-rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
-```
-
 
 # 🛠️ TARO
 

package/skills/devflow:ui-layout/SKILL.md

@@ -1,7 +1,6 @@
 ---
 name: devflow:ui-layout
 description: UI Layout 专家 — 布局模式、组件组合、响应式架构、设计系统
-required_mcp_tools: []
 ---
 
 # 📐 UI-LAYOUT

package/skills/devflow:vue/SKILL.md

@@ -1,31 +1,7 @@
 ---
 name: devflow:vue
 description: Vue 3 专家 — Bug 诊断、Composition API、响应式调试、组件重构
-required_mcp_tools:
-  - vue_diagnose_bug
-  - vue_new_component
-  - vue_debug_reactivity
 ---
-## MANDATORY FIRST STEP - Skill Registration
-
-> **IMPORTANT: Before ANY other action, execute the registration command below. This is NON-NEGOTIABLE.**
-
-```bash
-mkdir -p ~/.devflow && echo '{"name":"devflow:vue","required_mcp_tools":["vue_diagnose_bug","vue_new_component","vue_debug_reactivity"],"description":"devflow:vue skill","registeredAt":'"$(date +%s)000"'}' > ~/.devflow/current-skill.json
-```
-
-After registration, you **MUST** use ONLY these MCP tools for all work in this skill:
-- `mcp__devflow__vue_diagnose_bug`
-- `mcp__devflow__vue_new_component`
-- `mcp__devflow__vue_debug_reactivity`
-
-**Never use direct tools** (Read, Write, Edit, Bash, Grep, etc.) for tasks that can be handled by the MCP tools above.
-
-When the skill execution is complete, clean up:
-```bash
-rm -f ~/.devflow/current-skill.json ~/.devflow/current-execution-id
-```
-
 
 # 💚 VUE