ClawHub

Brand Voice Architect

Security checks across malware telemetry and agentic risk

Overview

This is a coherent brand-voice analysis and prompt-generation skill, with some transparency cautions around broad activation and generated prompts that silently rewrite prohibited terms.

Install this only if you want brand-voice analysis and prompt generation. Provide only writing samples you intend to analyze, and review generated system prompts before using them in another assistant, especially where exact wording, quotations, legal text, or compliance-sensitive language matters.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • System Prompt LeakageDirect Leakage, Indirect Extraction, Tool-Based Exfiltration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (3)

Lp3

Medium
Category
MCP Least Privilege
Confidence
93% confidence
Finding
The skill explicitly directs the agent to use local scripts and reference files, which implies file-read/code-adjacent capability, but it declares no permissions or boundaries. This creates a trust and enforcement gap: an orchestrator may allow the skill to access repository files or corpora without clear user-visible authorization, increasing the chance of unintended data exposure during analysis.

Vague Triggers

Medium
Confidence
89% confidence
Finding
The trigger guidance is overly expansive, including vague requests like making writing more consistent or asking what tone to use. That can cause the skill to activate in situations where the user did not intend a corpus-analysis or prompt-synthesis workflow, potentially pulling in files, generating system prompts, or shaping outputs beyond the user's expectation.

Natural-Language Policy Violations

Medium
Confidence
91% confidence
Finding
The generated system prompt explicitly instructs the downstream model to silently replace prohibited user-provided words with preferred equivalents. That creates undisclosed transformation of user input, which can misrepresent the user's meaning, hide sensitive or safety-relevant terms, and reduce transparency in audit or compliance contexts. In a brand-voice skill, this is more dangerous because the tool is designed to rewrite content at scale, making covert semantic drift likely and hard to detect.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal