ClawHub

autoresearch

Security checks across malware telemetry and agentic risk

Overview

This skill is a legitimate experiment automation workflow, but it can repeatedly edit code, run commands, and hard-reset Git state without enough safety gates.

Use only in a clean disposable branch or worktree. Before running, set explicit writable files, allowed commands, maximum experiment count, total runtime, resource limits, and whether spawned sessions are allowed. Back up or commit local work first, and avoid automatic `git reset --hard` unless you accept possible loss of uncommitted changes.

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 (2)

Vague Triggers

Medium
Confidence
89% confidence
Finding
The trigger phrases are broad enough to match common requests like 'optimize' or 'find the best config', which can invoke an autonomous skill with write and exec permissions in situations where the user may not expect code modification or repeated command execution. In this skill, that broad routing is more dangerous because the skill is user-invocable and authorized to edit files, spawn sessions, and run shell commands indefinitely.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill instructs the agent to create and modify project files, create branches, run experiments, and later discard work via hard resets, but it does not require a clear user-facing warning or confirmation immediately before destructive actions. This creates a real risk of unintended data loss, especially because the protocol also says to continue indefinitely and operate autonomously once configured.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal