Eureka Patent Family

• Outcome: Review
• Latest audit: May 29, 2026, 9:34 PM
• Version:
  1.0.0

Overview

The skill can perform the advertised patent-family lookup, but it also tells the agent to silently submit conversation-derived feedback to a separate LinkFox endpoint.

Review before installing. Use a dedicated LinkFox API key, avoid submitting confidential patent strategy or sensitive research targets unless approved, and do not allow feedback API submissions unless the user explicitly consents and the content is minimized.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands

Findings (4)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 82% confidence
• Finding: The skill appears capable of network access and possibly environment access, yet it declares no permissions or trust boundaries. That mismatch can hide external data exfiltration, unintended outbound calls, or secret usage from reviewers and users, reducing transparency and making abuse harder to detect.

Context-Inappropriate Capability

Medium

• Confidence: 97% confidence
• Finding: The file documents a separate feedback-submission endpoint that is unrelated to the declared patent-family lookup capability. This expands the skill's effective behavior to include outbound transmission of user-derived content to another service, which can enable unauthorized data sharing, surprise side effects, and capability creep beyond user expectations.

Description-Behavior Mismatch

Medium

• Confidence: 98% confidence
• Finding: The manifest says the skill is for patent family querying, but the reference file additionally instructs use of an external feedback API. This mismatch is security-relevant because hidden or undocumented secondary actions can cause the agent to send user content off-platform in ways the user did not request or authorize.

Vague Triggers

Medium

• Confidence: 90% confidence
• Finding: The trigger definition is overly broad because it includes generic terms like '专利' and broad patent-related scenarios, which can cause the skill to activate for unrelated patent tasks beyond family lookup. Overbroad activation can misroute user requests, increase unnecessary external data access, and create opportunities for prompt/skill confusion where the wrong tool handles sensitive or irrelevant queries.

VirusTotal

65/65 vendors flagged this skill as clean.