ClawHub
Back to skill

Security audit

MinerU OCR Local & API

Security checks across malware telemetry and agentic risk

Overview

This MinerU OCR skill is purpose-aligned, but users should understand that API mode can upload documents to a hosted service and save OCR outputs locally.

Install this only if you are comfortable choosing between hosted and local OCR. Use --mode local for confidential PDFs or images. Use API mode only when uploading the document to MinerU is acceptable, keep MINERU_API_TOKEN and MINERU_API_BASE_URL trusted, and delete saved OCR artifacts when they are no longer needed.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (3)

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The description explicitly advertises local-file upload to a hosted OCR API but does not warn that document contents may leave the local machine and be processed by a remote third party. This can cause accidental disclosure of sensitive PDFs, contracts, IDs, or proprietary documents when users assume OCR is local by default.

Missing User Warnings

Medium
Confidence
97% confidence
Finding
The concrete example for hosted API parsing from a local file normalizes a potentially risky action without stating that the file will be transmitted off-device. Because the example is copy-paste ready, it materially increases the chance that an agent or user will exfiltrate sensitive local documents unintentionally.

Missing User Warnings

Medium
Confidence
87% confidence
Finding
In API mode, the library uploads local documents to remote MinerU-managed endpoints and later downloads processed results, but the code provides no explicit user-facing consent, warning, or destination restriction. In an agent skill context, this can cause sensitive local files to be exfiltrated to third-party infrastructure without the operator clearly realizing it.

VirusTotal

60/60 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.