Mp Caveman

Security checks across malware telemetry and agentic risk

Overview

This skill only changes the assistant's writing style to a terse mode, with no code execution, file access, credential use, or hidden data flow.

Install this only if you want the assistant to keep using terse, compressed replies after activation. Remember the exact stop phrases are "stop caveman" and "normal mode"; use them if the style becomes unhelpful.

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: 93% confidence
Finding: The activation text uses broad trigger phrases such as '简短回复', '省token', '简洁', and '高效沟通', which can match many ordinary requests rather than an explicit opt-in. That creates unintended skill activation risk, causing response-style and policy behavior to change without clear user intent, which can degrade safety-critical communication or override expected assistant behavior.

Vague Triggers

Medium

Confidence: 95% confidence
Finding: The persistence rule says the mode remains active for every response once triggered and only turns off on specific phrases, making the behavior overly sticky and insufficiently bounded. Persistent style takeover can interfere with later unrelated tasks, reduce clarity in sensitive contexts, and create a durable instruction conflict that the user may not realize is still in effect.

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal