Biomarker Investigation Zhcn

• Outcome: Pass
• Latest audit: May 29, 2026 at 8:44 PM UTC
• Version:
  1.0.4

Overview

This skill is a disclosed PatSnap life-science research helper that sends biomarker and patent queries to a configured MCP service, with no evidence of hidden local access or destructive behavior.

Install only if you intend to use PatSnap's remote life-science MCP service. Use a dedicated, revocable PatSnap API key, avoid sharing MCP configuration or logs that contain the key-bearing URL, and treat patent-risk conclusions as research support rather than legal advice.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (4)

Description-Behavior Mismatch

Medium

• Confidence: 95% confidence
• Finding: The skill's declared purpose is limited to searching biomarker-related academic and patent literature, but the instructions expand the agent's role into patent infringement risk assessment. This scope expansion can cause the agent to handle higher-risk legal-analysis tasks that users and platform policy may not expect, increasing the chance of over-collection, overreach, or unsafe decision support.

Description-Behavior Mismatch

Medium

• Confidence: 94% confidence
• Finding: The skill description promises literature and patent searching, but the body authorizes use of broader datasets such as news, company, drug deal, clinical trial, and organization data. This creates a capability mismatch that can lead to unexpected data access and broader profiling or intelligence gathering than the user reasonably consented to.

Context-Inappropriate Capability

Medium

• Confidence: 93% confidence
• Finding: Adding patent infringement risk assessment goes beyond a straightforward biomarker literature search and introduces legal-risk analysis not necessary for the stated task. That expansion raises the danger of unauthorized legal-style guidance, especially if the agent produces conclusions users may rely on for business or compliance decisions.

Context-Inappropriate Capability

Low

• Confidence: 82% confidence
• Finding: The instructions permit generic web searching after MCP searches, but that capability is not disclosed in the skill declaration. While not inherently malicious, undisclosed external search broadens data egress and retrieval behavior beyond what users may expect from a literature-search skill.

VirusTotal

60/60 vendors flagged this skill as clean.