Security audit

industry-chain-intelligence

Security checks across malware telemetry and agentic risk

Overview

The skill is a local HTML report generator, but its bundled generator appears to produce hard-coded business intelligence while presenting the output as real-time Zhihuiya MCP data.

Review carefully before installing. Use it only for mockups or report-format drafting unless the publisher adds a real data-ingestion path, clear provenance checks, explicit user confirmation for MCP calls and file generation, and visible labeling when data is unavailable or estimated.

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 Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands

Findings (6)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 80% confidence
Finding: The skill explicitly instructs Phase 3 to call `scripts/generate_report.py` and describes producing an HTML artifact, which implies file creation/write capability without any declared permission boundary or explicit user-consent gate. In an agent environment, undeclared file output can surprise users, create persistent artifacts, and increase the blast radius if later extended to write outside an expected workspace.

Tp4

High

Category: MCP Tool Poisoning
Confidence: 92% confidence
Finding: The skill claims real-time Zhihuiya MCP-backed analysis and strict report fidelity, but the finding indicates the behavior may instead use static or fabricated content and undocumented report modes. This is dangerous because decision-makers may trust materially inaccurate outputs as data-backed intelligence, leading to bad strategic, procurement, or compliance decisions under false provenance.

Description-Behavior Mismatch

Medium

Confidence: 98% confidence
Finding: The skill metadata claims real-time MCP-driven industry intelligence, but this file generates reports entirely from hard-coded placeholder content and never retrieves live data. In a decision-support skill for CEOs and strategy teams, this is dangerous because users may treat fabricated or stale placeholders as authoritative analysis, leading to business decisions based on false premises.

Description-Behavior Mismatch

Medium

Confidence: 95% confidence
Finding: Government-only content is always rendered because the conditional uses `or True`, so enterprise-mode reports include sections intended for a different audience. This can cause information integrity and authorization issues by exposing content that users did not request and that may be inappropriate or misleading in enterprise contexts.

Vague Triggers

Medium

Confidence: 77% confidence
Finding: The trigger conditions include broad phrases and even a standalone company name interpreted via context, which can cause accidental activation. In a skill that performs external data retrieval and report generation, misfires can lead to unintended data access, unnecessary processing, and unwanted artifact creation without clear user intent.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: The skill description does not clearly warn users that it may send query terms to external MCP services and generate a local HTML file. Missing notice and consent around data egress and file output is risky because company names, strategic interests, or internal targets may be sensitive, and users may not realize their request causes external processing and persistent output.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.