AI-written code built log resource paths with path.join and no traversal check, letting remote attackers escape the log directory to list and read arbitrary files.

How AI contributed: Direct introductionPublished Apr 30, 2026 in zachhandley/zmcptools

Show full advisory description

Path traversal in the MCP Log Resource Handler (src/managers/ResourceManager.ts): resource templates logs://{dirname}/files and logs://{dirname}/content build paths via path.join(homedir(), '.mcptools', 'logs', dirname[, filename]) with no traversal containment check, so an attacker-controlled dirname (and the file query param) can escape ~/.mcptools/logs to list (readdir/stat) and read (readFile) arbitrary files (e.g. ../../../../etc/passwd). Exposed over MCP resources; remote exploitation possible; CVSS 6.3.

Root cause

Injection & unsafe execution introduced

Swap ScrapeJobEntry ick for Websites/WebsitePages, add Logs to Resources

AI-assisted change: Claude

793b9f5c44

src/managers/ResourceManager.ts src/services/AgentService.ts src/server/McpServer.ts

Fix status

No public fix identified

No minimum fix commit has been established for this finding.

Fix status unresolved

Case facts

CVE ID: CVE-2026-7445
GHSA ID: GHSA-6XQM-JW5J-72JF
Published: Apr 30, 2026
Severity: Not scored
Language: TypeScript
AI tool: Claude
AI role: Direct introduction
Vulnerable release: <= 0.2.2

Advisory source

AI-assisted change

AI-assisted changesrc/managers/ResourceManager.ts source zachhandley/zmcptools@793b9f5c44+27−6

@@ -6,7 +6,12 @@ import { WebScrapingService } from '../services/WebScrapingService.js'; import { DocumentationService } from '../services/DocumentationService.js'; import { MemoryService } from '../services/MemoryService.js'; import { VectorSearchService } from '../services/VectorSearchService.js';+import { WebsiteRepository } from '../repositories/WebsiteRepository.js';+import { WebsitePagesRepository } from '../repositories/WebsitePagesRepository.js'; import { PathUtils } from '../utils/pathUtils.js';+import { readdir, stat, readFile } from 'fs/promises';

Show 82 hidden lines

+import { join } from 'path';+import { homedir } from 'os';  export interface ResourceInfo {   uri: string;@@ -23,6 +28,8 @@ export class ResourceManager {   private documentationService: DocumentationService;   private memoryService: MemoryService;   private vectorSearchService: VectorSearchService;+  private websiteRepository: WebsiteRepository;+  private websitePagesRepository: WebsitePagesRepository;    constructor(     private db: DatabaseManager,@@ -42,6 +49,8 @@ export class ResourceManager {     );     this.documentationService = new DocumentationService(this.db);     this.vectorSearchService = new VectorSearchService(this.db);+    this.websiteRepository = new WebsiteRepository(this.db);+    this.websitePagesRepository = new WebsitePagesRepository(this.db);   }    /**@@ -52,13 +61,13 @@ export class ResourceManager {       {         uri: 'agents://list',         name: 'Agent List',-        description: 'List of all active agents with their status and metadata',+        description: 'List of all active agents with their status and metadata (use ?limit=50&offset=0&status=active&type=backend)',         mimeType: 'application/json'       },       {         uri: 'communication://rooms',         name: 'Communication Rooms',-        description: 'List of all active communication rooms',+        description: 'List of all active communication rooms (use ?limit=50&offset=0&search=text)',         mimeType: 'application/json'       },       {@@ -70,25 +79,37 @@ export class ResourceManager {       {         uri: 'scraping://jobs',         name: 'Scraper Jobs',-        description: 'List of web scraping jobs and their status',+        description: 'List of web scraping jobs and their status (use ?limit=50&offset=0&status=active&search=text)',         mimeType: 'application/json'       },       {         uri: 'documentation://sources',         name: 'Documentation Sources',-        description: 'List of scraped documentation sources',+        description: 'List of scraped documentation sources (use ?limit=50&offset=0&sourceType=api&search=text)',+        mimeType: 'application/json'+      },+      {+        uri: 'documentation://websites',+        name: 'Documentation Websites',+        description: 'List of all scraped websites (use ?limit=50&offset=0&search=text)',+        mimeType: 'application/json'+      },+      {+        uri: 'documentation://*/pages',+        name: 'Website Pages',+        description: 'List of pages for a specific website (use documentation://{websiteId}/pages?limit=50&offset=0&search=text)',         mimeType: 'application/json'       },       {         uri: 'agents://insights',         name: 'Agent Insights',-        description: 'Aggregated insights and learnings from agents',+        description: 'Aggregated insights and learnings from agents (use ?limit=100&offset=0&memoryType=insight&agentId=id&search=text)',         mimeType: 'application/json'       },       {         uri: 'vector://collections',         name: 'Vector Collections',-        description: 'List of ChromaDB vector collections and their statistics',+        description: 'List of ChromaDB vector collections and their statistics (use ?limit=50&offset=0&search=text)',         mimeType: 'application/json'       },       {@@ -102,6 +123,24 @@ export class ResourceManager {

Brings direct readdir, stat, and readFile primitives into the resource manager and adds website resource descriptors, extending the same object that later serves log paths to MCP clients.

AI-assisted changesrc/services/AgentService.ts source zachhandley/zmcptools@793b9f5c44+60−2

@@ -7,6 +7,7 @@ import type { ClaudeSpawnConfig } from '../process/ClaudeSpawner.js'; import { Logger } from '../utils/logger.js'; import { AgentPermissionManager } from '../utils/agentPermissions.js'; import { PathUtils } from '../utils/pathUtils.js';+import { FoundationCacheService } from './FoundationCacheService.js'; import { eq, and } from 'drizzle-orm'; import { agentSessions } from '../schemas/index.js'; import { resolve } from 'path';

Show 82 hidden lines

@@ -388,10 +389,36 @@ export class AgentService {       disallowedTools = agent.toolPermissions.disallowedTools;     } +    // 🚀 AUTOMATIC FOUNDATION CACHING INTEGRATION+    // Get or create foundation session for 85-90% token cost reduction+    let foundationSessionId: string | undefined;+    +    if (this.shouldUseFoundationCaching(agent)) {+      try {+        const cacheService = new FoundationCacheService(this.db);+        foundationSessionId = await cacheService.getOrCreateFoundationSession(+          workingDirectory,+          this.generateAgentContext(agent, taskDescription)+        );+        +        this.logger.info('Foundation session auto-created for agent', {+          agentId: agent.id,+          foundationSessionId,+          repositoryPath: workingDirectory,+          tokenSavings: '85-90% expected'+        });+      } catch (error) {+        this.logger.warn('Failed to create foundation session, continuing without caching', {+          agentId: agent.id,+          error: error instanceof Error ? error.message : String(error)+        });+      }+    }+     const config: ClaudeSpawnConfig = {       workingDirectory,       prompt,-      sessionId: `agent_${agent.id}`,+      sessionId: foundationSessionId || `agent_${agent.id}`, // Use foundation session if available       capabilities: agent.capabilities,       allowedTools: allowedTools,       disallowedTools: disallowedTools,@@ -401,7 +428,8 @@ export class AgentService {         AGENT_TYPE: agent.agentType || 'general_agent',         TASK_DESCRIPTION: taskDescription,         REPOSITORY_PATH: workingDirectory,-        ROOM_ID: agent.roomId || ''+        ROOM_ID: agent.roomId || '',+        FOUNDATION_SESSION_ID: foundationSessionId || ''       },       ...claudeConfig     };@@ -749,4 +777,36 @@ Focus on successfully completing your assigned task using your specialized capab       checkCompletion();     });   }++  // 🚀 FOUNDATION CACHING HELPER METHODS+  +  /**+   * Determines if an agent should use foundation caching+   * Enables for all agents by default, with opt-out capability+   */+  private shouldUseFoundationCaching(agent: AgentSession): boolean {+    // Allow opt-out via metadata+    if (agent.agentMetadata?.disableFoundationCaching === true) {+      return false;+    }+    +    // Enable for all agents by default+    return true;+  }+  +  /**+   * Generates context for foundation caching session+   */+  private generateAgentContext(agent: AgentSession, taskDescription: string): any {+    return {+      agentId: agent.id,+      agentName: agent.agentName,+      agentType: agent.agentType,+      taskDescription,+      capabilities: agent.capabilities,+      repositoryPath: agent.repositoryPath,+      timestamp: new Date().toISOString(),+      toolPermissions: agent.toolPermissions

Automatically creates and reuses a foundation caching session for agents, then propagates its identifier into the spawned process environment while adding unrelated runtime state changes to the same release.

AI-assisted changesrc/server/McpServer.ts source zachhandley/zmcptools@793b9f5c44+6−11

@@ -26,7 +26,7 @@ import { BrowserMcpTools } from '../tools/BrowserMcpTools.js'; import { WebScrapingMcpTools } from '../tools/WebScrapingMcpTools.js'; import { AnalysisMcpTools } from '../tools/AnalysisMcpTools.js'; import { TreeSummaryTools } from '../tools/TreeSummaryTools.js';-import { CacheMcpTools } from '../tools/CacheMcpTools.js';+// CacheMcpTools removed - foundation caching now automatic import { BrowserTools } from '../tools/BrowserTools.js'; import { WebScrapingService } from '../services/WebScrapingService.js';

Show 44 hidden lines

 import { AgentService, MemoryService, FileOperationsService, TreeSummaryService, fileOperationsService } from '../services/index.js';@@ -51,7 +51,7 @@ export class McpServer {   private webScrapingMcpTools: WebScrapingMcpTools;   private analysisMcpTools: AnalysisMcpTools;   private treeSummaryTools: TreeSummaryTools;-  private cacheMcpTools: CacheMcpTools;+  // cacheMcpTools removed - foundation caching now automatic   private fileOperationsService: FileOperationsService;   private treeSummaryService: TreeSummaryService;   private resourceManager: ResourceManager;@@ -113,7 +113,7 @@ export class McpServer {     this.webScrapingMcpTools = new WebScrapingMcpTools(webScrapingService, memoryService, this.repositoryPath);     this.analysisMcpTools = new AnalysisMcpTools(memoryService, this.repositoryPath);     this.treeSummaryTools = new TreeSummaryTools();-    this.cacheMcpTools = new CacheMcpTools(this.db);+    // Foundation caching is now automatic - no manual tools needed      // Initialize managers     this.resourceManager = new ResourceManager(this.db, this.repositoryPath);@@ -249,9 +249,8 @@ export class McpServer {       // Analysis and file operation tools       ...this.analysisMcpTools.getTools(),       // TreeSummary tools-      ...this.treeSummaryTools.getTools(),-      // Foundation Cache tools-      ...this.cacheMcpTools.getTools()+      ...this.treeSummaryTools.getTools()+      // Foundation Cache tools removed - now automatic     ];   } @@ -652,11 +651,7 @@ export class McpServer {       return await this.treeSummaryTools.handleToolCall(name, args);     } -    // Check Foundation Cache tools-    const cacheToolNames = this.cacheMcpTools.getTools().map(t => t.name);-    if (cacheToolNames.includes(name)) {-      return await this.cacheMcpTools.handleToolCall(name, args);-    }+    // Foundation Cache tools removed - now automatic      // Handle orchestration tools     switch (name) {

Removes manual cache tool registration and dispatch handling because foundation caching is automatic, leaving resource manager requests as the still-exposed MCP surface in this server.

AI-written code built log resource paths with path.join and no traversal check, letting remote attackers escape the log directory to list and read arbitrary files.

How AI contributed: Direct introductionPublished Apr 30, 2026 in

zachhandley/zmcptools

Show full advisory description

AI-assisted change

AI-assisted changesrc/managers/ResourceManager.ts source zachhandley/zmcptools@793b9f5c44+27−6

@@ -6,7 +6,12 @@ import { WebScrapingService } from '../services/WebScrapingService.js'; import { DocumentationService } from '../services/DocumentationService.js'; import { MemoryService } from '../services/MemoryService.js'; import { VectorSearchService } from '../services/VectorSearchService.js';+import { WebsiteRepository } from '../repositories/WebsiteRepository.js';+import { WebsitePagesRepository } from '../repositories/WebsitePagesRepository.js'; import { PathUtils } from '../utils/pathUtils.js';+import { readdir, stat, readFile } from 'fs/promises';

Show 82 hidden lines

+import { join } from 'path';+import { homedir } from 'os';  export interface ResourceInfo {   uri: string;@@ -23,6 +28,8 @@ export class ResourceManager {   private documentationService: DocumentationService;   private memoryService: MemoryService;   private vectorSearchService: VectorSearchService;+  private websiteRepository: WebsiteRepository;+  private websitePagesRepository: WebsitePagesRepository;    constructor(     private db: DatabaseManager,@@ -42,6 +49,8 @@ export class ResourceManager {     );     this.documentationService = new DocumentationService(this.db);     this.vectorSearchService = new VectorSearchService(this.db);+    this.websiteRepository = new WebsiteRepository(this.db);+    this.websitePagesRepository = new WebsitePagesRepository(this.db);   }    /**@@ -52,13 +61,13 @@ export class ResourceManager {       {         uri: 'agents://list',         name: 'Agent List',-        description: 'List of all active agents with their status and metadata',+        description: 'List of all active agents with their status and metadata (use ?limit=50&offset=0&status=active&type=backend)',         mimeType: 'application/json'       },       {         uri: 'communication://rooms',         name: 'Communication Rooms',-        description: 'List of all active communication rooms',+        description: 'List of all active communication rooms (use ?limit=50&offset=0&search=text)',         mimeType: 'application/json'       },       {@@ -70,25 +79,37 @@ export class ResourceManager {       {         uri: 'scraping://jobs',         name: 'Scraper Jobs',-        description: 'List of web scraping jobs and their status',+        description: 'List of web scraping jobs and their status (use ?limit=50&offset=0&status=active&search=text)',         mimeType: 'application/json'       },       {         uri: 'documentation://sources',         name: 'Documentation Sources',-        description: 'List of scraped documentation sources',+        description: 'List of scraped documentation sources (use ?limit=50&offset=0&sourceType=api&search=text)',+        mimeType: 'application/json'+      },+      {+        uri: 'documentation://websites',+        name: 'Documentation Websites',+        description: 'List of all scraped websites (use ?limit=50&offset=0&search=text)',+        mimeType: 'application/json'+      },+      {+        uri: 'documentation://*/pages',+        name: 'Website Pages',+        description: 'List of pages for a specific website (use documentation://{websiteId}/pages?limit=50&offset=0&search=text)',         mimeType: 'application/json'       },       {         uri: 'agents://insights',         name: 'Agent Insights',-        description: 'Aggregated insights and learnings from agents',+        description: 'Aggregated insights and learnings from agents (use ?limit=100&offset=0&memoryType=insight&agentId=id&search=text)',         mimeType: 'application/json'       },       {         uri: 'vector://collections',         name: 'Vector Collections',-        description: 'List of ChromaDB vector collections and their statistics',+        description: 'List of ChromaDB vector collections and their statistics (use ?limit=50&offset=0&search=text)',         mimeType: 'application/json'       },       {@@ -102,6 +123,24 @@ export class ResourceManager {

Brings direct readdir, stat, and readFile primitives into the resource manager and adds website resource descriptors, extending the same object that later serves log paths to MCP clients.

AI-assisted changesrc/services/AgentService.ts source zachhandley/zmcptools@793b9f5c44+60−2

@@ -7,6 +7,7 @@ import type { ClaudeSpawnConfig } from '../process/ClaudeSpawner.js'; import { Logger } from '../utils/logger.js'; import { AgentPermissionManager } from '../utils/agentPermissions.js'; import { PathUtils } from '../utils/pathUtils.js';+import { FoundationCacheService } from './FoundationCacheService.js'; import { eq, and } from 'drizzle-orm'; import { agentSessions } from '../schemas/index.js'; import { resolve } from 'path';

Show 82 hidden lines

@@ -388,10 +389,36 @@ export class AgentService {       disallowedTools = agent.toolPermissions.disallowedTools;     } +    // 🚀 AUTOMATIC FOUNDATION CACHING INTEGRATION+    // Get or create foundation session for 85-90% token cost reduction+    let foundationSessionId: string | undefined;+    +    if (this.shouldUseFoundationCaching(agent)) {+      try {+        const cacheService = new FoundationCacheService(this.db);+        foundationSessionId = await cacheService.getOrCreateFoundationSession(+          workingDirectory,+          this.generateAgentContext(agent, taskDescription)+        );+        +        this.logger.info('Foundation session auto-created for agent', {+          agentId: agent.id,+          foundationSessionId,+          repositoryPath: workingDirectory,+          tokenSavings: '85-90% expected'+        });+      } catch (error) {+        this.logger.warn('Failed to create foundation session, continuing without caching', {+          agentId: agent.id,+          error: error instanceof Error ? error.message : String(error)+        });+      }+    }+     const config: ClaudeSpawnConfig = {       workingDirectory,       prompt,-      sessionId: `agent_${agent.id}`,+      sessionId: foundationSessionId || `agent_${agent.id}`, // Use foundation session if available       capabilities: agent.capabilities,       allowedTools: allowedTools,       disallowedTools: disallowedTools,@@ -401,7 +428,8 @@ export class AgentService {         AGENT_TYPE: agent.agentType || 'general_agent',         TASK_DESCRIPTION: taskDescription,         REPOSITORY_PATH: workingDirectory,-        ROOM_ID: agent.roomId || ''+        ROOM_ID: agent.roomId || '',+        FOUNDATION_SESSION_ID: foundationSessionId || ''       },       ...claudeConfig     };@@ -749,4 +777,36 @@ Focus on successfully completing your assigned task using your specialized capab       checkCompletion();     });   }++  // 🚀 FOUNDATION CACHING HELPER METHODS+  +  /**+   * Determines if an agent should use foundation caching+   * Enables for all agents by default, with opt-out capability+   */+  private shouldUseFoundationCaching(agent: AgentSession): boolean {+    // Allow opt-out via metadata+    if (agent.agentMetadata?.disableFoundationCaching === true) {+      return false;+    }+    +    // Enable for all agents by default+    return true;+  }+  +  /**+   * Generates context for foundation caching session+   */+  private generateAgentContext(agent: AgentSession, taskDescription: string): any {+    return {+      agentId: agent.id,+      agentName: agent.agentName,+      agentType: agent.agentType,+      taskDescription,+      capabilities: agent.capabilities,+      repositoryPath: agent.repositoryPath,+      timestamp: new Date().toISOString(),+      toolPermissions: agent.toolPermissions

AI-assisted changesrc/server/McpServer.ts source zachhandley/zmcptools@793b9f5c44+6−11

@@ -26,7 +26,7 @@ import { BrowserMcpTools } from '../tools/BrowserMcpTools.js'; import { WebScrapingMcpTools } from '../tools/WebScrapingMcpTools.js'; import { AnalysisMcpTools } from '../tools/AnalysisMcpTools.js'; import { TreeSummaryTools } from '../tools/TreeSummaryTools.js';-import { CacheMcpTools } from '../tools/CacheMcpTools.js';+// CacheMcpTools removed - foundation caching now automatic import { BrowserTools } from '../tools/BrowserTools.js'; import { WebScrapingService } from '../services/WebScrapingService.js';

Show 44 hidden lines

 import { AgentService, MemoryService, FileOperationsService, TreeSummaryService, fileOperationsService } from '../services/index.js';@@ -51,7 +51,7 @@ export class McpServer {   private webScrapingMcpTools: WebScrapingMcpTools;   private analysisMcpTools: AnalysisMcpTools;   private treeSummaryTools: TreeSummaryTools;-  private cacheMcpTools: CacheMcpTools;+  // cacheMcpTools removed - foundation caching now automatic   private fileOperationsService: FileOperationsService;   private treeSummaryService: TreeSummaryService;   private resourceManager: ResourceManager;@@ -113,7 +113,7 @@ export class McpServer {     this.webScrapingMcpTools = new WebScrapingMcpTools(webScrapingService, memoryService, this.repositoryPath);     this.analysisMcpTools = new AnalysisMcpTools(memoryService, this.repositoryPath);     this.treeSummaryTools = new TreeSummaryTools();-    this.cacheMcpTools = new CacheMcpTools(this.db);+    // Foundation caching is now automatic - no manual tools needed      // Initialize managers     this.resourceManager = new ResourceManager(this.db, this.repositoryPath);@@ -249,9 +249,8 @@ export class McpServer {       // Analysis and file operation tools       ...this.analysisMcpTools.getTools(),       // TreeSummary tools-      ...this.treeSummaryTools.getTools(),-      // Foundation Cache tools-      ...this.cacheMcpTools.getTools()+      ...this.treeSummaryTools.getTools()+      // Foundation Cache tools removed - now automatic     ];   } @@ -652,11 +651,7 @@ export class McpServer {       return await this.treeSummaryTools.handleToolCall(name, args);     } -    // Check Foundation Cache tools-    const cacheToolNames = this.cacheMcpTools.getTools().map(t => t.name);-    if (cacheToolNames.includes(name)) {-      return await this.cacheMcpTools.handleToolCall(name, args);-    }+    // Foundation Cache tools removed - now automatic      // Handle orchestration tools     switch (name) {

Removes manual cache tool registration and dispatch handling because foundation caching is automatic, leaving resource manager requests as the still-exposed MCP surface in this server.