ClawHub
Back to skill

Security audit

identify-patent-white-space-ip

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed patent white-space analysis workflow with user confirmation gates and proportionate optional patent-data retrieval.

Installers should expect this skill to guide a patent white-space analysis, ask for confirmation at two decision points, optionally use PatSnap MCP tools only when authorized, and create a local HTML report that may include sensitive patent-map or business-context details.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
Findings (3)

Description-Behavior Mismatch

Medium
Confidence
86% confidence
Finding
The skill requires explicit user confirmation at two checkpoints, but then instructs the agent to automatically generate and write an HTML report after Stage 7 without a separate request. This creates an instruction-flow inconsistency that can cause unrequested side effects, especially file creation in environments where artifact generation should be user-driven and tightly scoped.

Context-Inappropriate Capability

Medium
Confidence
83% confidence
Finding
The skill expands from analysis into workspace and file-management behavior by directing creation of a self-contained HTML report in a 'safe workspace directory' and assuming path handling logic. That introduces unnecessary file I/O capabilities relative to the stated analytical purpose, increasing the attack surface for unwanted persistence, data leakage into artifacts, or misuse of writable tool access.

Description-Behavior Mismatch

High
Confidence
90% confidence
Finding
The skill states that it should not perform downstream technical validation, yet it authorizes live MCP use for 'bottom-level patent validation or additional retrieval.' In a patent-analysis context, this contradiction can cause the agent to cross from hypothesis generation into external retrieval and validation work that the skill explicitly says must remain excluded, risking scope creep, unauthorized network access, and misleading users about what has been validated.

VirusTotal

61/61 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.