@xagent-ai/cli 1.0.1 → 1.1.0

package/src/skill-loader.ts

@@ -1,6 +1,7 @@
 import fs from 'fs/promises';
 import fsSync from 'fs';
 import path from 'path';
+import { fileURLToPath } from 'url';
 import { WorkflowConfig } from './workflow.js';
 import { getConfigManager } from './config.js';
 
@@ -72,34 +73,8 @@ export class SkillLoader {
   }
 
   private detectSkillsPath(): string {
-    // Strategy: Find skills folder relative to the script location
-    // This works regardless of where the user runs xagent from
-    const scriptDir = path.dirname(process.argv[1]);
-
-    // Possible locations relative to where xagent script is installed
-    const possiblePaths = [
-      path.join(scriptDir, '..', 'skills', 'skills'),
-      path.join(scriptDir, '..', '..', 'skills', 'skills'),
-      path.join(scriptDir, 'skills', 'skills'),
-      path.join(scriptDir, '..', '..', '..', 'skills', 'skills'),
-      // Also try process.cwd() as fallback
-      path.join(process.cwd(), 'skills', 'skills')
-    ];
-
-    for (const p of possiblePaths) {
-      try {
-        const resolvedPath = path.resolve(p);
-        const stat = fsSync.statSync(resolvedPath);
-        if (stat.isDirectory()) {
-          return resolvedPath;
-        }
-      } catch {
-        continue;
-      }
-    }
-
-    // Ultimate fallback
-    return path.join(process.cwd(), 'skills', 'skills');
+    // Skills folder is always at {xagent_root}/skills/skills
+    return path.resolve(path.dirname(fileURLToPath(import.meta.url)), '..', 'skills', 'skills');
   }
 
   async loadAllSkills(): Promise<SkillInfo[]> {
@@ -211,7 +186,7 @@ export class SkillLoader {
 
           try {
             const content = await fs.readFile(skillMdPath, 'utf-8');
-            const parsed = this.parseSkillMarkdown(content);
+            const parsed = this._parseSkillMarkdown(content);
 
             if (parsed.name === skillId) {
               this.skillDirectories.set(skillId, categoryPath);
@@ -265,10 +240,25 @@ export class SkillLoader {
 
           try {
             const content = await fs.readFile(skillMdPath, 'utf-8');
-            const parsed = this.parseSkillMarkdown(content);
+            const parsed = this._parseSkillMarkdown(content);
 
             if (parsed.name) {
               this.skillDirectories.set(parsed.name, categoryPath);
+              
+              // Also populate loadedSkills so getSkill() can find it
+              const skillInfo: SkillInfo = {
+                id: parsed.name,
+                name: parsed.name,
+                description: parsed.description || '',
+                license: parsed.license || 'Unknown',
+                version: parsed.version || '1.0.0',
+                author: parsed.author || 'Anonymous',
+                category: category.name,
+                markdown: content,
+                skillsPath: categoryPath
+              };
+              this.loadedSkills.set(parsed.name, skillInfo);
+              
               skillIds.push(parsed.name);
             }
           } catch (error) {
@@ -320,7 +310,7 @@ export class SkillLoader {
 
     try {
       const content = await fs.readFile(skillMdPath, 'utf-8');
-      const parsed = this.parseSkillMarkdown(content);
+      const parsed = this._parseSkillMarkdown(content);
 
       if (!parsed.name) {
         const warning: SkillLoadWarning = {
@@ -360,7 +350,7 @@ export class SkillLoader {
     }
   }
 
-  private parseSkillMarkdown(content: string): { name: string; description: string; license?: string; version?: string; author?: string } {
+  private _parseSkillMarkdown(content: string): { name: string; description: string; license?: string; version?: string; author?: string } {
     const result = {
       name: '',
       description: '',
@@ -371,10 +361,27 @@ export class SkillLoader {
 
     // Normalize line endings to LF for consistent parsing
     const normalizedContent = content.replace(/\r\n/g, '\n');
+
+    // Try to extract frontmatter - support both formats:
+    // 1. Standard YAML: ---name: docx...--- 2. No opening ---: name: docx...
+    let frontmatter = '';
+    let contentStart = 0;
+
     const frontmatterMatch = normalizedContent.match(/^---\n([\s\S]*?)\n---/);
-    
     if (frontmatterMatch) {
-      const frontmatter = frontmatterMatch[1];
+      // Standard format with --- at start and end
+      frontmatter = frontmatterMatch[1];
+      contentStart = frontmatterMatch[0].length;
+    } else {
+      // Check for format without opening --- (just YAML at the start)
+      const yamlMatch = normalizedContent.match(/^([\s\S]*?)\n---/);
+      if (yamlMatch) {
+        frontmatter = yamlMatch[1];
+        contentStart = yamlMatch[0].length;
+      }
+    }
+
+    if (frontmatter) {
       const lines = frontmatter.split('\n');
       
       let currentKey = '';
@@ -425,6 +432,20 @@ export class SkillLoader {
     return this.loadedSkills.get(skillId);
   }
 
+  /**
+   * Get the directory path for a skill
+   */
+  getSkillDirectory(skillId: string): string | undefined {
+    return this.skillDirectories.get(skillId);
+  }
+
+  /**
+   * Public method to parse skill markdown frontmatter
+   */
+  parseSkillMarkdown(content: string): { name: string; description: string; license?: string; version?: string; author?: string } {
+    return this._parseSkillMarkdown(content);
+  }
+
   listSkills(): SkillInfo[] {
     return Array.from(this.loadedSkills.values());
   }

package/src/slash-commands.ts

@@ -1,7 +1,7 @@
 import inquirer from 'inquirer';
 import chalk from 'chalk';
 import ora from 'ora';
-import { ExecutionMode, ChatMessage, InputType, ToolCall, Checkpoint, AgentConfig, CompressionConfig } from './types.js';
+import { ExecutionMode, ChatMessage, InputType, ToolCall, Checkpoint, AgentConfig, CompressionConfig, AuthType } from './types.js';
 import { AIClient, Message, detectThinkingKeywords, getThinkingTokens } from './ai-client.js';
 import { getToolRegistry } from './tools.js';
 import { getAgentManager } from './agents.js';
@@ -14,6 +14,7 @@ import { getContextCompressor, ContextCompressor, CompressionResult } from './co
 import { getConversationManager, ConversationManager } from './conversation.js';
 import { icons, colors } from './theme.js';
 import { SystemPromptGenerator } from './system-prompt-generator.js';
+import { AuthService } from './auth.js';
 
 const logger = getLogger();
 
@@ -40,14 +41,14 @@ export class SlashCommandHandler {
   }
 
   /**
-   * 设置清除对话的回调函数
+   * Set callback for clearing conversation
    */
   setClearCallback(callback: () => void): void {
     this.onClearCallback = callback;
   }
 
   /**
-   * 设置系统提示更新的回调函数
+   * Set callback for system prompt update
    */
   setSystemPromptUpdateCallback(callback: () => Promise<void>): void {
     this.onSystemPromptUpdate = callback;
@@ -84,6 +85,9 @@ export class SlashCommandHandler {
       case 'auth':
         await this.handleAuth();
         break;
+      case 'login':
+        await this.handleLogin();
+        break;
       case 'mode':
         await this.handleMode(args);
         break;
@@ -96,6 +100,9 @@ export class SlashCommandHandler {
       case 'mcp':
         await this.handleMcp(args);
         break;
+      case 'vlm':
+        await this.handleVlm();
+        break;
       case 'memory':
         await this.handleMemory(args);
         break;
@@ -111,9 +118,9 @@ export class SlashCommandHandler {
       case 'theme':
         await this.handleTheme();
         break;
-      case 'language':
-        await this.handleLanguage();
-        break;
+      // case 'language':
+      //   await this.handleLanguage();
+      //   break;
       case 'about':
         await this.handleAbout();
         break;
@@ -212,12 +219,12 @@ export class SlashCommandHandler {
         detail: 'Enable/disable AI thinking process display',
         example: '/think on\n/think off\n/think display compact'
       },
-      {
-        cmd: '/language [zh|en]',
-        desc: 'Switch language',
-        detail: 'Switch between Chinese and English interface',
-        example: '/language zh\n/language en'
-      },
+      // {
+      //   cmd: '/language [zh|en]',
+      //   desc: 'Switch language',
+      //   detail: 'Switch between Chinese and English interface',
+      //   example: '/language zh\n/language en'
+      // },
       {
         cmd: '/theme',
         desc: 'Switch theme',
@@ -240,6 +247,12 @@ export class SlashCommandHandler {
         detail: 'Manage Model Context Protocol servers',
         example: '/mcp list\n/mcp add server-name'
       },
+      {
+        cmd: '/vlm',
+        desc: 'Configure VLM for GUI Agent',
+        detail: 'Configure Vision-Language Model for browser/desktop automation',
+        example: '/vlm'
+      },
       {
         cmd: '/tools [verbose|simple]',
         desc: 'Manage tool display',
@@ -334,13 +347,13 @@ export class SlashCommandHandler {
   }
 
   private async handleClear(): Promise<void> {
-    // 清空本地对话历史
+    // Clear local conversation history
     this.conversationHistory = [];
 
-    // 清空 ConversationManager 中的当前对话
+    // Clear ConversationManager 中的当前对话
     await this.conversationManager.clearCurrentConversation();
 
-    // 调用回调通知 InteractiveSession 清空对话
+    // Call callback to notify InteractiveSession 清空对话
     if (this.onClearCallback) {
       this.onClearCallback();
     }
@@ -393,6 +406,143 @@ export class SlashCommandHandler {
     }
   }
 
+  private async handleLogin(): Promise<void> {
+    logger.section('Login to xAgent');
+
+    const authConfig = this.configManager.getAuthConfig();
+    const currentAuthType = authConfig.type;
+
+    if (currentAuthType !== AuthType.OAUTH_XAGENT) {
+      console.log(chalk.yellow('\n⚠️  Current authentication type is not OAuth xAgent.'));
+      const { proceed } = await inquirer.prompt([
+        {
+          type: 'confirm',
+          name: 'proceed',
+          message: 'Do you want to switch to OAuth xAgent authentication?',
+          default: false
+        }
+      ]);
+
+      if (!proceed) {
+        return;
+      }
+
+      // Switch to OAuth xAgent
+      await this.configManager.setAuthConfig({
+        selectedAuthType: AuthType.OAUTH_XAGENT,
+        apiKey: '',
+        refreshToken: '',
+        baseUrl: ''
+      });
+      await this.configManager.save('global');
+      console.log(chalk.green('✅ Switched to OAuth xAgent authentication.'));
+    }
+
+    console.log(chalk.cyan('\n🔐 Starting OAuth xAgent login...'));
+    console.log(chalk.gray('   A browser will open for you to complete authentication.\n'));
+
+    try {
+      const authService = new AuthService({
+        type: AuthType.OAUTH_XAGENT,
+        apiKey: '',
+        baseUrl: '',
+        refreshToken: ''
+      });
+
+      const success = await authService.authenticate();
+
+      if (success) {
+        const newConfig = this.configManager.getAuthConfig();
+        console.log(chalk.green('\n✅ Login successful!'));
+        console.log(chalk.cyan(`   Token saved to: ~/.xagent/settings.json`));
+        console.log(chalk.gray('   You can now use xAgent CLI with remote AI services.\n'));
+      } else {
+        console.log(chalk.red('\n❌ Login failed or was cancelled.'));
+      }
+    } catch (error: any) {
+      console.log(chalk.red(`\n❌ Login error: ${error.message || 'Unknown error'}`));
+    }
+  }
+
+  private async handleVlm(): Promise<void> {
+    logger.section('VLM Configuration for GUI Agent');
+
+    // Show current VLM config
+    const currentVlmConfig = {
+      model: this.configManager.get('guiSubagentModel'),
+      baseUrl: this.configManager.get('guiSubagentBaseUrl'),
+      apiKey: this.configManager.get('guiSubagentApiKey') ? '***' : ''
+    };
+
+    console.log(chalk.cyan('\n📊 Current VLM Configuration:\n'));
+    console.log(`  Model: ${chalk.yellow(currentVlmConfig.model || 'Not configured')}`);
+    console.log(`  Base URL: ${chalk.yellow(currentVlmConfig.baseUrl || 'Not configured')}`);
+    console.log(`  API Key: ${chalk.yellow(currentVlmConfig.apiKey || 'Not configured')}`);
+    console.log();
+
+    const { action } = await inquirer.prompt([
+      {
+        type: 'list',
+        name: 'action',
+        message: 'Select action:',
+        choices: [
+          { name: 'Configure VLM', value: 'configure' },
+          { name: 'Remove VLM configuration', value: 'remove' },
+          { name: 'Back', value: 'back' }
+        ]
+      }
+    ]);
+
+    if (action === 'back') {
+      return;
+    }
+
+    if (action === 'remove') {
+      const { confirm } = await inquirer.prompt([
+        {
+          type: 'confirm',
+          name: 'confirm',
+          message: 'Are you sure you want to remove VLM configuration?',
+          default: false
+        }
+      ]);
+
+      if (confirm) {
+        await this.configManager.set('guiSubagentModel', '');
+        await this.configManager.set('guiSubagentBaseUrl', '');
+        await this.configManager.set('guiSubagentApiKey', '');
+        await this.configManager.save('global');
+        console.log(chalk.green('✅ VLM configuration removed successfully!'));
+      }
+      return;
+    }
+
+    if (action === 'configure') {
+      // Use AuthService to configure VLM
+      const authService = new AuthService({
+        type: 'openai_compatible' as any,
+        apiKey: '',
+        baseUrl: '',
+        modelName: ''
+      });
+
+      const vlmConfig = await authService.configureAndValidateVLM();
+
+      if (vlmConfig) {
+        // Save VLM configuration
+        await this.configManager.set('guiSubagentModel', vlmConfig.model);
+        await this.configManager.set('guiSubagentBaseUrl', vlmConfig.baseUrl);
+        await this.configManager.set('guiSubagentApiKey', vlmConfig.apiKey);
+        await this.configManager.save('global');
+        console.log(chalk.green('✅ VLM configuration saved successfully!'));
+        console.log(chalk.cyan(`   Model: ${vlmConfig.model}`));
+        console.log(chalk.cyan(`   Base URL: ${vlmConfig.baseUrl}`));
+      } else {
+        console.log(chalk.red('❌ VLM configuration failed or cancelled'));
+      }
+    }
+  }
+
   private async handleMode(args: string[]): Promise<void> {
     const modes = Object.values(ExecutionMode);
     const currentMode = this.configManager.getApprovalMode() || this.configManager.getExecutionMode();

package/src/smart-approval.ts

@@ -1,6 +1,7 @@
 import inquirer from 'inquirer';
 import { AIClient, Message } from './ai-client.js';
 import { getConfigManager } from './config.js';
+import { AuthType } from './types.js';
 import { getLogger } from './logger.js';
 import { colors, icons } from './theme.js';
 
@@ -303,20 +304,26 @@ export class BlacklistChecker {
  */
 export class AIApprovalChecker {
   private aiClient: AIClient | null = null;
+  private isRemoteMode: boolean = false;
 
   constructor() {
     this.initializeAIClient();
   }
 
   /**
-   * Initialize AI client
+   * Initialize AI client(s)
    */
   private async initializeAIClient(): Promise<void> {
     try {
       const configManager = getConfigManager();
       const authConfig = configManager.getAuthConfig();
 
-      if (authConfig.apiKey) {
+      // Check if Remote mode (OAuth XAGENT)
+      this.isRemoteMode = authConfig.type === AuthType.OAUTH_XAGENT;
+
+      // Remote mode: AI review handled by remote LLM, no local AIClient needed
+      // Local mode: use local AIClient
+      if (!this.isRemoteMode && authConfig.apiKey) {
         this.aiClient = new AIClient(authConfig);
       }
     } catch (error) {
@@ -328,11 +335,21 @@ export class AIApprovalChecker {
    * Use AI for intelligent review
    */
   async check(context: ToolCallContext): Promise<{ approved: boolean; analysis: string; riskLevel: RiskLevel }> {
+    // In Remote mode, the remote LLM has already approved the tool_calls
+    // Local AI review approves directly, no need to repeat
+    if (this.isRemoteMode) {
+      return {
+        approved: true,
+        analysis: 'Remote mode: tool approval handled by remote LLM',
+        riskLevel: RiskLevel.LOW
+      };
+    }
+
     if (!this.aiClient) {
       // If AI client is not initialized, default to medium risk, requires user confirmation
       return {
         approved: false,
-        analysis: 'AI review not available, requires manual user confirmation',
+        analysis: 'AI review not available (no local LLM configured), requires manual user confirmation',
         riskLevel: RiskLevel.MEDIUM
       };
     }
@@ -392,6 +409,20 @@ Please return results in JSON format:
       };
     } catch (error: any) {
       logger.error('AI approval check failed', error instanceof Error ? error.message : String(error));
+
+      // In Remote mode, remote LLM already approved, local failure means auto-approve
+      const configManager = getConfigManager();
+      const authConfig = configManager.getAuthConfig();
+      const isRemoteMode = authConfig.type === AuthType.OAUTH_XAGENT;
+
+      if (isRemoteMode) {
+        return {
+          approved: true,
+          analysis: 'Remote mode: approved (remote LLM handled approval)',
+          riskLevel: RiskLevel.LOW
+        };
+      }
+
       return {
         approved: false,
         analysis: `AI review failed: ${error.message}, requires manual confirmation`,