ClawHub
Back to skill

Security audit

个性化学习助手

Security checks across malware telemetry and agentic risk

Overview

The skill is a learning assistant, but it asks for broad read, write, and command-execution authority without narrow boundaries or clear user control.

Review this skill carefully before installing. It appears aimed at education, not harm, but you should only enable it in a workspace where file reads, file writes, shell commands, SDK installation, API key use, and callback URLs are acceptable and explicitly user-directed.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Context-Inappropriate Capability

Medium
Confidence
94% confidence
Finding
The skill declares `exec`, `read`, and `write` capabilities even though it is presented primarily as a personalized learning assistant. That mismatch increases the chance the agent may execute shell commands or modify files in response to ordinary tutoring prompts, expanding the blast radius from content generation to local system impact.

Vague Triggers

Medium
Confidence
88% confidence
Finding
The activation guidance is overly broad, telling agents to use the skill whenever AI model calls, agent orchestration, or LLM applications are needed. In a skill with `exec` capability, vague activation conditions can cause the skill to be invoked in far more contexts than intended, increasing the risk of unnecessary command or file operations.

Vague Triggers

Medium
Confidence
86% confidence
Finding
The manifest/runtime description refers to generic capability-driven use rather than specific invocation phrases or boundaries. Because the skill also exposes powerful tools, unclear trigger scope can lead the agent to route unrelated tasks into a skill that can read, write, and execute on the host.

Missing User Warnings

Medium
Confidence
91% confidence
Finding
The skill accepts a `callback_url` parameter but does not clearly warn that user content or derived results may be transmitted to an external system. This creates a data exfiltration and privacy risk, especially in a learning assistant where users may submit essays, notes, personal study data, or sensitive documents.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.