Security audit

PDF和图片文字提取

Security checks across malware telemetry and agentic risk

Overview

This is a straightforward PDF and image text extraction skill, with no evidence of hidden network access, credential use, or destructive behavior.

Install if you need local text extraction from PDFs or images. Avoid using highly sensitive documents unless you are comfortable with the extracted text being shown in chat and, if you ask to save it, persisted as a Markdown file that you may need to delete later.

SkillSpector

By NVIDIA

Vulnerability Patterns

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
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (5)

Description-Behavior Mismatch

Medium

Confidence: 94% confidence
Finding: The skill states that processed files are not stored, yet the workflow explicitly includes generating and saving Markdown output files containing extracted document text. This creates a privacy and data-handling discrepancy that can mislead users into sharing sensitive documents under false assumptions about retention.

Intent-Code Divergence

Medium

Confidence: 95% confidence
Finding: The privacy note says files will not be stored, but elsewhere the skill instructs the agent to save extracted content to Markdown files. Because extracted text may contain sensitive personal or business information, this inconsistency can lead to unintended retention and exposure of confidential data.

Vague Triggers

Medium

Confidence: 90% confidence
Finding: The README says users can invoke the skill with broad natural-language requests like 'extract the text' or 'convert this PDF to text' without defining clearer trigger boundaries or requiring an attached image/PDF. In an agent setting, this can cause over-triggering on ambiguous prompts and unintended processing of files or content, which is a real security and safety concern even if not overtly malicious.

Missing User Warnings

Low

Confidence: 80% confidence
Finding: The README states 'Save the extracted text' will generate a Markdown file, but it does not clearly warn users up front that a new file will be created as a side effect. Undisclosed file creation is a legitimate safety issue because users may not expect persistence of extracted content, which can increase privacy, data handling, and workspace integrity risks.

Missing User Warnings

Medium

Confidence: 77% confidence
Finding: The README states that users can ask to save extracted results, but it does not clearly warn that extracted text may be written into a Markdown file, which can create confidentiality and data-handling risks. OCR output often contains sensitive information from contracts, IDs, reports, or research documents, so silent persistence increases the chance of unintended retention, disclosure, or downstream sharing.

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.