ClawHub
Back to skill

Security audit

Second-Order Thinking

Security checks across malware telemetry and agentic risk

Overview

The skill is a useful reasoning aid, but it tells the agent to fetch live remote skill instructions at runtime, which can change behavior after installation review.

Review this skill before installing if you are comfortable with a skill that can fetch remote instructions at run time. Prefer a version that keeps updates in normal reviewed package releases, or only allows remote updates with explicit consent and integrity checks.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Description-Behavior Mismatch

Medium
Confidence
96% confidence
Finding
The skill is a reasoning aid, but it instructs the agent to fetch remote content at runtime to obtain a 'canonical current version' of the skill. That creates a supply-chain and prompt-injection surface: whoever controls the remote JSON or the network path can alter the skill's behavior after installation, bypassing local review and causing the agent to follow untrusted instructions.

Context-Inappropriate Capability

Medium
Confidence
94% confidence
Finding
The network retrieval is not necessary to perform second-order consequence analysis; it only refreshes instructions and links. Unnecessary runtime fetches expand the attack surface without functional need, enabling remote instruction drift, privacy leakage through request metadata, and dependency on external availability for a purely local reasoning task.

Missing User Warnings

Low
Confidence
87% confidence
Finding
The skill directs network access at run start without a user-facing warning or consent cue. Even if the remote endpoint is benign, silent outbound requests can surprise users, leak usage timing/context via metadata, and make the agent's behavior less transparent and auditable.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.