Security audit

Clawhub

Security checks across malware telemetry and agentic risk

Overview

This skill is a coherent Android automation connector, but it grants broad live device-control power and understates the privacy and state-changing risks.

Install only if you intend to let OpenClaw control connected Android devices. Prefer test devices, confirm the device serial before actions, avoid sensitive apps or personal data on-screen, use multi-device execution only when you explicitly want it, and revoke or disable USB debugging when finished.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
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 (6)

Intent-Code Divergence

Medium

Confidence: 96% confidence
Finding: The safety section materially understates the tool's capabilities: the documented commands can capture screenshots, inspect UI trees, enumerate installed apps, input text, launch apps, and control multiple devices. This can expose sensitive information on-screen and perform state-changing actions, so claiming it will not access or modify sensitive data may mislead users and downstream agents into taking insufficient precautions.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The README advertises direct Android control, app launching, text input, and concurrent multi-device actions, but does not warn users that these capabilities can immediately change device state and affect every connected phone. In the context of an AI-controlled automation skill, omission of safety warnings increases the chance of unintended taps, app launches, or bulk actions causing data loss, privacy issues, or operational disruption.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The installation steps tell users to enable Developer Options and USB debugging, but do not explain that this grants powerful device-control access to the host and any tool using ADB. Because this skill is specifically designed to automate devices through MCP and ADB, failing to warn about the security implications can expose connected phones to unauthorized control, data access, or accidental modification if the host environment is compromised or misused.

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: The examples encourage natural-language commands like opening apps, tapping coordinates, and typing text without clarifying that these commands will directly execute on the connected device. In an AI-agent context, this is risky because users may treat examples as harmless queries rather than live device actions, leading to accidental interaction with sensitive apps, messages, or settings.

Vague Triggers

Medium

Confidence: 89% confidence
Finding: The trigger phrases are very broad, everyday commands like taking screenshots, clicking, opening apps, and sending messages, with no explicit gating around consent, device selection, or action sensitivity. In an agent environment, this increases the chance of unintended activation or overbroad execution on a connected phone, especially when multiple devices may be present.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The skill documents powerful device-control operations but does not prominently warn users that it can change device state, interact with apps, input arbitrary text, navigate screens, and execute concurrent actions across multiple phones. Without a clear warning, users and orchestrating agents may underestimate the risk of destructive, privacy-impacting, or bulk actions.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.