Security audit

夸克扫描王 OCR文字识别 - yescan ocr universal

Security checks across malware telemetry and agentic risk

Overview

This OCR skill sends user-provided images to Quark's OCR API as disclosed, with no evidence of hidden persistence, credential theft, destructive behavior, or unrelated execution.

Install only if you are comfortable sending provided images, including IDs, invoices, or medical documents, to Quark's OCR service using your API key. Avoid submitting documents whose contents should not leave your machine.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
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 (5)

Tp4

High

Category: MCP Tool Poisoning
Confidence: 84% confidence
Finding: The documentation presents the skill as OCR-only, but the referenced behavior includes broader image transformation, cleanup, document conversion, and local file output capabilities. Undisclosed file-writing and image-manipulation features expand the attack surface, can surprise users, and may enable unintended processing or persistence of sensitive documents beyond simple OCR.

Vague Triggers

Medium

Confidence: 75% confidence
Finding: The fallback intent 'extract pure text' is overly broad and may match ambiguous everyday requests, causing the skill to send user-supplied images to the remote OCR service when the user did not clearly intend that action. In this skill, that matters because uploaded images can contain highly sensitive identity, medical, or financial information and are transmitted to a third party.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The code reads arbitrary local files, base64-encodes their contents, and sends them to an external OCR API. In an OCR skill that handles sensitive documents such as IDs, licenses, invoices, and medical reports, this creates a real data exfiltration/privacy risk if users are not clearly informed that local content leaves the host and is processed by a third-party service.

Missing User Warnings

Medium

Confidence: 73% confidence
Finding: The function writes OCR-derived image content to local disk without any indication in this module that the action is disclosed, consented to, or controllable by the user. In an OCR skill that may process IDs, invoices, medical reports, or other sensitive documents, silent persistence can create privacy and data-retention risk if files remain on shared hosts or are later accessed by other processes or users.

Missing User Warnings

Medium

Confidence: 78% confidence
Finding: This logic persists generated Word/Excel/PDF files to disk without any visible user-facing warning or consent path in the module. Because the skill handles potentially highly sensitive source material such as identity documents, financial records, and medical reports, silent local storage increases exposure through residual files, accidental sharing, or collection by other software on the same system.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.