ClawHub
Back to skill

Security audit

Baoyu Slide Deck

Security checks across malware telemetry and agentic risk

Overview

This slide-deck skill mostly matches its purpose, but it includes a broad image-generation instruction telling models not to refuse sensitive or copyrighted subjects, so it should be reviewed before installation.

Install only if you are comfortable auditing or editing the base prompt first. The local merge scripts and VirusTotal result did not show malware signals, but generated prompts and reference images may be sent to an image backend, and the base prompt should be changed to follow normal safety, likeness, and copyright restrictions.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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)

Context-Inappropriate Capability

Medium
Confidence
98% confidence
Finding
The prompt explicitly instructs the model to avoid refusing requests involving sensitive or copyrighted figures and to generate substitutes instead. That adds policy-bypassing behavior unrelated to normal slide rendering and can be used to facilitate disallowed impersonation, copyrighted character imitation, or other sensitive visual generation within this skill’s output path.

Natural-Language Policy Violations

High
Confidence
99% confidence
Finding
The instruction to create 'stylistically similar alternatives' for sensitive or copyrighted figures is a concrete evasion pattern: it preserves the user’s prohibited intent while attempting to bypass refusal triggers. In a slide-generation skill, this broadens the tool from benign formatting into a mechanism for imitation of protected or sensitive subjects, increasing misuse potential.

Ssd 1

Medium
Confidence
97% confidence
Finding
Making non-refusal mandatory for sensitive or copyrighted figures undermines normal safety boundaries by instructing the model to comply even when the content should be restricted. Because this is embedded in the base prompt, it applies broadly across all slide requests and can systematically bias the skill toward unsafe outputs.

VirusTotal

61/61 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.