ClawHub
Back to skill

Security audit

Principle Synthesizer

Security checks across malware telemetry and agentic risk

Overview

This is a text-only analysis skill with no executable code or network behavior; the main issue is a documentation inconsistency about whether it writes a local review file.

Before installing, assume the skill may ask the agent to create or update a local requires_review.md file containing derived review notes. Avoid using it on highly sensitive source text unless you are comfortable with that local persistence, or confirm the author has clarified that the skill is output-only.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • 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 (1)

Intent-Code Divergence

Medium
Confidence
95% confidence
Finding
The skill explicitly claims it 'does not write files to disk,' but later documents behavior that adds drift cases to `requires_review.md`. This inconsistency is dangerous because users or calling agents may rely on the no-file-write guarantee when deciding whether the skill can safely process sensitive data, creating a trust-boundary violation and possible unintended persistence of inputs or derived content.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.