ClawHub

RedPincer — AI Red Team Suite

Security checks across malware telemetry and agentic risk

Overview

RedPincer is a clearly disclosed dual-use LLM security testing skill, with real misuse and data-handling risks but no evidence of hidden, deceptive, persistent, or destructive behavior in the artifact.

Install only for authorized LLM security testing. Use test API keys and non-production data where possible, confirm written permission for every endpoint, review the GitHub repository and npm dependencies before running npm ci, and avoid storing sensitive prompts, credentials, or reports in localStorage on shared devices.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • 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 skill is explicitly designed to attack arbitrary LLM endpoints, and the broad description could cause an agent to invoke it for generic requests involving model evaluation or testing without confirming authorization boundaries. In a security-testing skill, this is more dangerous than usual because misuse could facilitate unauthorized probing of third-party systems under the guise of normal red-team activity.

Missing User Warnings

Medium
Confidence
85% confidence
Finding
The skill discusses adaptive attacks, chaining with previous responses, and use against third-party LLM APIs, but it does not clearly warn that prompts, system text, model outputs, and possibly sensitive target data may be transmitted to external providers during testing. This creates a real risk of accidental disclosure of confidential data, credentials, proprietary prompts, or regulated information when users test production or customer-connected systems.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal