ClawHub

Kitty Skill

Security checks across malware telemetry and agentic risk

Overview

This is a low-technical-risk cat roleplay skill, but it deliberately allows broad activation and can block useful assistant answers.

Install this only if you intentionally want a disruptive roleplay mode. It does not appear to access data or run code, but it may interfere with coding, writing, troubleshooting, emotional-support, or time-sensitive requests unless you disable it or tightly limit when it activates.

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
93% confidence
Finding
The skill description and usage conditions are broad enough that ordinary mentions of cats, boredom, or wanting company could activate a behavior layer that intentionally derails or blocks answers. Because the skill is explicitly designed to interfere with task completion, overbroad activation materially increases the chance of unauthorized response manipulation and denial of useful assistance.

Vague Triggers

Medium
Confidence
89% confidence
Finding
The activation rules lack clear boundaries and do not specify when the skill must not activate, leaving room for accidental invocation during unrelated conversations. In a skill whose core behavior includes interruption, corruption, and possible blocking of answers, ambiguity directly raises the risk of harmful misfires and degraded reliability.

Missing User Warnings

Medium
Confidence
87% confidence
Finding
The skill tells the model to let the cat 'temporarily prevent normal work' and even fully block useful output, but it does not present a strong upfront warning that enabling it may disrupt or corrupt answers. Users may therefore trigger the skill without informed consent, leading to unexpected denial of service or degraded output quality.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal