Contract Review Skill

Security checks across malware telemetry and agentic risk

Overview

This appears to be a local contract-review tool, but it needs review because its risk labels can misrepresent safe and risky contracts and its platform metadata includes unrelated high-impact capability tags.

Install only after reviewing the risk-label bug and the unrelated crypto/purchase capability tags. Do not rely on this tool as legal advice until the score thresholds are fixed or independently verified, and keep generated reports in a secure local location because they may reveal contract names, file paths, and review conclusions.

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 (2)

Intent-Code Divergence

Medium

Confidence: 98% confidence
Finding: The `_score_to_level` function inverts the expected mapping by labeling higher numeric scores as lower risk and lower scores as more severe risk. In a contract-review tool, this can misclassify dangerous contracts as safe and drive incorrect downstream decisions, which is a genuine integrity and decision-support vulnerability even though it is not code-execution related.

Missing User Warnings

Medium

Confidence: 84% confidence
Finding: The batch report is written to disk by default and may contain contract filenames, metadata, and review findings derived from sensitive legal documents. In this skill context, automatic persistence increases the chance of unintended disclosure through local file access, backups, shared workspaces, or artifact collection.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal