ClawHub
Back to skill

Security audit

mt5-httpapi

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed MT5 trading and backtesting client with real-money authority, but its sensitive actions are scoped to a user-provided endpoint and require explicit user confirmation.

Install only if you operate the mt5-httpapi server yourself and understand that it can affect a real brokerage account. Keep the API bound to localhost or protected by strong authentication, use demo accounts first, confirm every trade or terminal-control action carefully, and do not expose an unauthenticated instance to a network.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (2)

Context-Inappropriate Capability

Medium
Confidence
94% confidence
Finding
The backtest runbook explicitly tells the agent to verify local files under assets/experts and assets/sets, which expands the skill from a pure HTTP client into local workspace inspection. That conflicts with the earlier no-autonomous-file-reading safety model and can normalize broader filesystem access than the user may expect, especially in shared or sensitive repositories.

Intent-Code Divergence

High
Confidence
97% confidence
Finding
The skill says it must never read repository files autonomously for secrets, but later instructions direct the agent to inspect local workspace files for backtest assets. This inconsistent policy creates a dangerous exception path where an agent may generalize file-reading behavior and accidentally access sensitive repository contents such as configs, tokens, or other secrets while following the runbook.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

Detected: suspicious.exposed_secret_literal

File appears to expose a hardcoded API secret or token.

Critical
Code
suspicious.exposed_secret_literal
Location
SKILL.md:55