Stock Monitor Pro

• Outcome: Pass
• Latest audit: May 29, 2026, 8:14 PM
• Version:
  2.1.0

Overview

This is a disclosed stock-alert monitor that runs locally and polls public market-data sources, with some caution needed around its persistent background mode and scraping-style data access.

Install only if you want a local stock-monitoring daemon. Edit the watchlist, costs, and thresholds before starting it, expect your monitored symbols to be sent to external finance data providers, and use ./control.sh stop when you no longer want it running. Be mindful that the data-fetching approach may conflict with provider terms if run aggressively or through proxy-based workarounds.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration

Findings (5)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 86% confidence
• Finding: The skill describes and invokes network-capable behavior such as market data retrieval, anti-scraping session handling, and multi-source fetching, yet no permissions are declared. This creates a transparency and governance gap: users and the hosting platform cannot accurately assess or constrain outbound connectivity, which is especially risky for a background daemon that runs continuously.

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 93% confidence
• Finding: The documented purpose says the skill is for A-shares and ETFs monitoring, but the analysis indicates additional capabilities including news/sentiment retrieval, fund-flow and 龙虎榜 data collection, macro correlation logic, AI-style advisory output, and London gold monitoring. This mismatch is dangerous because it hides the real operational scope, expands data collection and decision-making behavior beyond user expectations, and can enable undisclosed external access or risky financial guidance.

Context-Inappropriate Capability

Medium

• Confidence: 93% confidence
• Finding: The code deliberately impersonates browser traffic with detailed fingerprint-like headers, randomized request timing, and fallback guidance to use a proxy/WARP when blocked. In an agent skill, this goes beyond normal API consumption and enables stealthy scraping behavior that can violate provider terms, evade rate limits, and be repurposed for abusive collection at scale.

Ssd 2

Medium

• Confidence: 95% confidence
• Finding: The comments and implementation explicitly frame the skill as using anti-detection scraping tactics, including browser impersonation and randomized human-like timing. In skill context, that is risky because it operationalizes evasive collection patterns that can be reused for unauthorized scraping and concealment of automated access.

Ssd 2

Medium

• Confidence: 96% confidence
• Finding: The skill includes explicit guidance to use a proxy/WARP or tune frequency when persistent blocking occurs, which is effectively operational advice for bypassing access controls. In an autonomous agent setting, such instructions increase the likelihood that the skill will be used to continue scraping after detection rather than failing safely.

VirusTotal

65/65 vendors flagged this skill as clean.