ClawHub
Back to skill

Security audit

Skill Auto-Use

Security checks across malware telemetry and agentic risk

Overview

The skill is transparent, but it would persistently make the agent auto-run other skills without asking, including potentially sensitive ones.

Review this before installing. It is not malicious code, but it changes agent behavior in a durable way. Use it only with a tightly reviewed trigger table, keep automatic triggers limited to low-risk read-only work, and require explicit approval before any skill touches files, accounts, public content, credentials, spending, or installed-skill configuration.

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

High
Confidence
97% confidence
Finding
The skill’s top-level description mandates automatic use of installed skills across session start, installs, context matches, and audits, creating a broadly self-propagating activation policy. This can cause unrelated or risky skills to be invoked without explicit user consent, expanding attack surface and enabling chained misuse if any installed skill is unsafe or overly privileged.

Vague Triggers

High
Confidence
99% confidence
Finding
The instruction to scan for trigger matches on every user message creates an always-on dispatcher that can reinterpret ordinary conversation as authorization to invoke tools or skills. In this skill’s context, that is especially dangerous because it suppresses user confirmation and can repeatedly trigger multiple skills from ambiguous language.

Vague Triggers

Medium
Confidence
92% confidence
Finding
The example triggers include broad categories such as research questions, corrections, keywords, domains, and task types that overlap heavily with normal conversation. Because the skill encourages automatic execution when triggers match, these loose patterns can cause frequent unintended activation of skills, including those that access external systems or modify persistent state.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.