ClawHub

ClawBrain Benchmark

Security checks across malware telemetry and agentic risk

Overview

This appears to be a benchmark skill, but it may exercise broad tool powers under fairly generic requests without enough user gating.

Install only if you intentionally want an agent benchmark that may run commands, use network requests, and exercise other tool workflows. Before using it, review the skill text for confirmation prompts, dry-run options, and limits on messaging or service-management actions; run it in a disposable or sandboxed workspace if those limits are not explicit.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (2)

Vague Triggers

Medium
Confidence
95% confidence
Finding
The trigger phrases are generic natural-language requests such as running a benchmark or testing model performance, which can easily appear in ordinary conversation and unintentionally invoke the skill. Because the skill is configured as user-invocable and dispatches to an exec tool, accidental activation could lead to benchmark actions that touch files, terminal commands, network access, or messaging-related workflows without clear user intent.

Missing User Warnings

Medium
Confidence
98% confidence
Finding
The skill advertises benchmarking across file operations, web fetching, messaging, and terminal/service management, but it does not warn the user that invoking the benchmark may exercise high-impact capabilities. In context, this is more dangerous because the skill uses command-dispatch to an exec tool and explicitly requires curl, increasing the likelihood of filesystem changes, command execution, and network activity occurring under a vague benchmark request.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal