ClawHub
Back to skill

Security audit

Claw Switch

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed model switcher, but it can change OpenClaw's active model and restart the gateway from broad natural-language triggers.

Install only if you want an always-available OpenClaw router that may switch models and restart the gateway based on normal task wording. Review provider costs and privacy expectations, and avoid using it during active work unless you are comfortable with possible session interruption.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Vague Triggers

Medium
Confidence
95% confidence
Finding
The trigger phrases are broad everyday terms like 图片, 文件, PDF, 截图, 视觉, and similar natural-language descriptions, so the skill may activate during ordinary conversation rather than an explicit request to switch models. Because this skill performs system-affecting actions, unintended activation can cause surprise model changes and follow-on gateway restarts that disrupt active sessions.

Vague Triggers

Medium
Confidence
91% confidence
Finding
The description says the router works with natural language and no manual config, which creates an ambiguous activation boundary and encourages matching on general task descriptions instead of explicit administrative intent. In a skill that changes runtime configuration, this increases the chance of accidental invocation and unauthorized or unintended environment changes.

Missing User Warnings

Medium
Confidence
97% confidence
Finding
The skill instructs an immediate 'openclaw gateway restart' as part of normal execution without warning that this is a disruptive system action. If triggered unexpectedly, it can interrupt active requests, break user workflows, and turn an accidental model-routing event into a service availability issue.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.