ClawHub
Back to skill

Security audit

presentation-generator

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed presentation-generation skill with optional PDF export and public sharing, so the main user risk is accidentally publishing confidential deck content.

Install this only if you are comfortable with Node/npm tooling, Playwright/Chromium for PDF export, and optional Vercel publishing. Do not run deploy unless the deck is approved for external visibility, and avoid deploying a whole directory that may contain non-deck files or secrets.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
Findings (5)

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding
The skill's declared purpose focuses on deck generation, but the content also instructs the agent to deploy generated artifacts to public Vercel URLs and invoke external tooling such as Playwright/Chromium. That mismatch matters because deployment is an externally visible side effect that can expose confidential presentation content or embedded local assets if a user or higher-level policy assumes this skill is generation-only.

Description-Behavior Mismatch

High
Confidence
95% confidence
Finding
This script performs a real external deployment to Vercel and returns a public URL, which creates a data-exfiltration/publication path for potentially confidential decks. In the context of a presentation-generation skill, this is a scope-expanding capability that can expose sensitive business content if triggered without explicit, informed user consent.

Context-Inappropriate Capability

Medium
Confidence
89% confidence
Finding
Bundling third-party hosting into a content-generation skill introduces an outbound network action that is not necessary for generating presentations and increases the blast radius of prompt or agent misuse. Even though the script warns about public visibility and uses preview by default, it still enables unauthorized publication of sensitive material to an external service.

Unpinned Dependencies

Low
Category
Supply Chain
Content
"private": true,
  "description": "Runtime-only deps for export-pdf.sh — installed on first run, not part of the published skill package.",
  "dependencies": {
    "playwright": "^1.46.0"
  }
}
Confidence
93% confidence
Finding
"playwright": "^1.46.0"

Known Vulnerable Dependency: playwright==1.46.0 — 1 advisory(ies): CVE-2025-59288 (Playwright downloads and installs browsers without verifying the authenticity of)

High
Category
Supply Chain
Confidence
97% confidence
Finding
playwright==1.46.0

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.