Back to skill

Security audit

Local MCP Server

Security checks for vulnerabilities and agentic risk

Overview

This skill is coherent for a local MCP server, but it gives AI tools local file and shell access with only limited scoping guidance.

Review this carefully before installing. Use it only in a dedicated sandbox or least-privilege account, restrict allowedPaths to specific project folders, keep bash disabled unless needed, and do not expose secrets, home directories, SSH keys, tokens, or private app data to models through this server.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The README advertises high-risk capabilities—reading files, executing bash commands, and searching the local filesystem—without an explicit warning that connecting an LLM to these tools can expose sensitive data or enable destructive command execution. In this skill's context, the danger is elevated because it is designed to give local models direct tool access on-device, and the stated protections are only briefly claimed rather than clearly bounded and cautioned.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill explicitly advertises bash command execution and local file access, but the documentation does not provide a prominent warning that these capabilities can modify the system, expose sensitive files, or execute destructive commands if invoked by an LLM or user workflow. In the context of an MCP server that exposes Read, Bash, Grep, and glob to local models, this materially increases risk because the advertised functionality enables direct interaction with the host environment.

Content

No source excerpt is available for this finding.