Security audit

QUANT-X v10 量化策略仪表盘

Security checks across malware telemetry and agentic risk

Overview

This is a stock-analysis dashboard with disclosed live market polling, but it overstates its quantitative methods and gives strong trading actions that users could mistake for rigorous financial advice.

Review this as an informational dashboard only. Before installing or deploying it, note that it sends watched stock symbols to Tencent market-data services every 3 seconds, loads an external charting library, and produces buy/sell-style signals from simplified heuristics rather than full technical-indicator or backtesting logic. Do not rely on its position-sizing recommendations without independent verification and appropriate financial advice.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (6)

Description-Behavior Mismatch

Medium

Confidence: 94% confidence
Finding: The HTML advertises advanced quantitative capabilities such as multifactor scoring, OBI dimensions, sector linkage, big-order tiering, and Backtrader-style analysis, but the implementation uses simplistic heuristics and fabricated derivations from a single quote feed. In a trading-analysis skill, misleading users about analytical rigor can directly influence financial decisions, making this more dangerous than ordinary marketing overstatement.

Intent-Code Divergence

Medium

Confidence: 95% confidence
Finding: The UI labels claim specific indicator families like MA/MACD, RSI/KDJ, OBV/OBI, BOLL/ATR, capital-flow, and sector linkage, but the code does not calculate these indicators and instead maps price change, amplitude, and order-volume ratios to canned labels. This is deceptive in a financial decision-support context because users may trust the dashboard as if it were performing established technical analysis when it is not.

Intent-Code Divergence

Medium

Confidence: 96% confidence
Finding: The classic strategies section claims concrete implementations of dual moving average, Bollinger grid trading, Donchian breakout, momentum, and Z-score mean reversion, but the code substitutes threshold checks on intraday percent change and amplitude. In a quant-trading skill, this mismatch can mislead users into acting on outputs they believe are strategy-backed signals, increasing the chance of financial harm.

Vague Triggers

Medium

Confidence: 88% confidence
Finding: The skill description advertises very broad financial-analysis and decision-support use cases without defining clear invocation boundaries, exclusions, or safety constraints. In an agent setting, this can cause over-activation in loosely related stock or trading conversations and increase the chance the model provides trading guidance or triggers network/data-fetch behaviors when the user did not explicitly request them.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The README instructs repeated external requests to Tencent market-data endpoints every 3 seconds and promotes public browser deployment, but provides no warning about outbound network activity, third-party data sharing, rate limits, or exposure of queried symbols and access patterns. In practice, this can leak user interest/watchlist data to external services and encourage unsafe deployment of a networked dashboard without informed consent.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The file gives highly prescriptive trading recommendations such as '满仓' and '清仓' tied to score thresholds without any balancing warning about financial risk, uncertainty, or suitability. In a skill explicitly designed for real-time stock analysis and buy/sell signaling, users could treat these outputs as actionable financial advice and suffer asset losses if the model, data, or formula is wrong or stale.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.