ClawHub
Back to skill

Security audit

Beta Lead Scoring

Security checks across malware telemetry and agentic risk

Overview

This skill appears to be a local CSV lead-scoring script, but it materially overstates itself as a LightGBM/SHAP AI model with API support.

Install only if you are comfortable treating this as a rule-based prototype, not a validated AI lead-scoring model. Do not rely on its scores for important sales, compliance, or procurement decisions unless the publisher corrects the documentation or ships the promised model, explainability, dependency, and API functionality.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Tp4

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding
The skill metadata and documentation claim AI-powered lead scoring with LightGBM, SHAP interpretability, and API integration, but the provided skill file shows only a documented local CSV workflow and no evidence of model loading, inference logic, SHAP generation, or API support. This is dangerous because users may make business decisions based on assumed model-backed output and may install dependencies or trust capabilities that do not actually exist, creating integrity and supply-chain trust issues even if there is no direct code execution in this file.

Description-Behavior Mismatch

Medium
Confidence
97% confidence
Finding
The skill advertises an AI-powered LightGBM + SHAP lead-scoring system, but the implementation is only a simple heuristic formula with no model loading, inference, or explainability logic. This is a trust and integrity vulnerability: downstream users may make business decisions under false assumptions about model quality, validation, and interpretability, which can lead to unsafe automation, compliance issues, or procurement deception.

Intent-Code Divergence

Medium
Confidence
96% confidence
Finding
The module documentation misrepresents the scoring logic as a predictive ML system while the code itself states it is a placeholder rule-based scorer. In this context, the mismatch is especially risky because lead scoring influences sales prioritization and resource allocation; users may incorrectly assume the outputs are model-derived and explainable, when they are not.

VirusTotal

62/62 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.