Security audit

Markdown转换器免费版

Security checks across malware telemetry and agentic risk

Overview

This document-conversion skill is mostly coherent, but it needs Review because broad activation wording and optional cloud/plugin processing are not clearly bounded for sensitive documents.

Install only if you are comfortable letting the agent run uvx/markitdown on chosen files. Avoid using the Azure -d/-e options or --use-plugins on confidential documents unless you have reviewed those providers and explicitly intend to send document content outside the local environment. Confirm output paths before conversion to avoid unwanted files or overwrites.

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)

Intent-Code Divergence

Medium

Confidence: 91% confidence
Finding: The skill claims basic conversion is local-only, yet elsewhere documents cloud-backed processing via Azure Document Intelligence and third-party plugins. This can mislead users into exposing sensitive document contents to external services without informed consent, especially in an agent context where execution may be automated.

Vague Triggers

High

Confidence: 93% confidence
Finding: Overly broad trigger keywords increase the chance that the skill activates for ordinary requests unrelated to safe document conversion. In an agent with `exec` and `write` permissions, accidental invocation can cause unintended command execution, file creation, or processing of sensitive local content.

Vague Triggers

Medium

Confidence: 90% confidence
Finding: The 'Use when' guidance is broad and underspecified, making activation scope ambiguous. This is dangerous because the skill has execution capability and may be selected in contexts where the user did not intend file processing, leading to unnecessary access to local files or command invocation.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: The examples show writing output files but do not clearly warn users that converted content will be saved to disk. In an agent setting, this can surprise users by leaving sensitive derived artifacts in the workspace or overwriting expected files if paths are reused.

Missing User Warnings

High

Confidence: 95% confidence
Finding: The documentation omits a clear warning that some modes may send document contents to external cloud services. Because converted inputs may contain sensitive business or personal data, silent use of Azure services or plugins can cause unintended data disclosure and compliance violations.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.