ClawHub
Back to skill

Security audit

Osh Contract Review

Security checks across malware telemetry and agentic risk

Overview

This is a plain-language contract review skill with no code execution or hidden access, though users should not treat its signing guidance as legal advice.

Use this skill to understand contract terms and identify questions or risks, but do not rely on it as a lawyer or final sign/no-sign decision maker. Avoid pasting highly sensitive contracts unless you are comfortable sharing that text with your agent environment, and consult a qualified lawyer for high-value, employment, IP, non-compete, or dispute-related contracts.

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)

Description-Behavior Mismatch

Medium
Confidence
93% confidence
Finding
The skill says it must not provide legal advice, but later requires an overall conclusion about whether to sign and under what conditions. In a contract-review context, that recommendation can reasonably be interpreted as legal guidance, creating unsafe reliance and exposing users to material decision-making based on unqualified advice.

Intent-Code Divergence

Medium
Confidence
95% confidence
Finding
The contradictory instructions create a policy gap: the assistant is told not to give legal advice but is also required to produce a signing recommendation, which effectively functions as advice. In practice, this increases the chance the model will cross from explanation into prescriptive legal judgment in a high-stakes domain.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.