ClawHub
Back to skill

Security audit

only for test

Security checks across malware telemetry and agentic risk

Overview

This skill is a coherent document-to-Markdown extractor, with important privacy cautions for OCR and temporary PDF images.

Install this only if you are comfortable with the Python packages and the configured MCP OCR tool. Use it on files you choose, and avoid OCR for confidential scans or images unless you understand that service's privacy and retention behavior; also check or clean /tmp after OCR on sensitive PDFs.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Description-Behavior Mismatch

Medium
Confidence
91% confidence
Finding
The skill is presented as read-only, but the PDF OCR workflow writes rendered page images to /tmp before sending them to the OCR tool. That discrepancy can violate user expectations, leave sensitive document remnants on disk, and weaken the claimed non-mutating security model even if the original source file is not modified.

Scope Creep

Medium
Confidence
97% confidence
Finding
Granting Write and Bash to a skill advertised as read-only extraction violates least privilege and expands the attack surface well beyond its stated purpose. If the skill is invoked on untrusted inputs or later modified, those permissions could enable filesystem changes, package installation, command execution, or data exfiltration paths unrelated to document reading.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
The OCR instructions allow image paths or URLs to be sent to an external MCP OCR service without any privacy or data-handling warning. In a document-extraction context, inputs are likely to contain sensitive business, personal, or regulated information, so silent transmission to another service increases confidentiality risk and user surprise.

VirusTotal

62/62 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.