Security audit

Word Docx V1 Tool Free

Security checks across malware telemetry and agentic risk

Overview

This is a Markdown-only Word document helper with proportionate file and command use, though its optional callback and network-dependency notes need user awareness.

Install only if you are comfortable with an agent creating or modifying local Word files and running Python package commands. Avoid using callback_url unless you intentionally want a remote endpoint notified, and choose output filenames carefully so existing documents are not overwritten.

SkillSpector

By NVIDIA

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Behavioral ASTexec() Call, eval() Call, Dynamic Import
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

Findings (6)

Description-Behavior Mismatch

Medium

Confidence: 90% confidence
Finding: The skill is presented as a local Word/docx editing tool, but it also documents callback URLs and network diagnostics that expand its operational scope into external communications. That mismatch can mislead users and agents about data flow, increasing the risk of unintended exfiltration or network use in a context where only local file processing is expected.

Context-Inappropriate Capability

Medium

Confidence: 92% confidence
Finding: Adding a callback_url to a basic personal document tool introduces outbound communication that is not justified by the stated functionality. If used, document contents, metadata, or task results could be sent to an external endpoint without users reasonably expecting network transmission.

Context-Inappropriate Capability

Medium

Confidence: 88% confidence
Finding: The documented installation of requests adds general-purpose network access that is unrelated to basic docx creation/editing. In an agent environment with exec capability, unnecessary network libraries broaden the attack surface and make covert external communication easier.

Intent-Code Divergence

Low

Confidence: 74% confidence
Finding: The documentation claims no extra API key or external service is needed, while other sections imply callback and network behavior. This inconsistency can cause users to underestimate privacy and security implications, reducing informed consent around external communications.

Missing User Warnings

Medium

Confidence: 86% confidence
Finding: The skill declares write and exec capabilities without clearly warning users about filesystem modification and command execution risks. In an agent context, understated privileged capabilities can lead to unsafe trust decisions and unintended local system changes.

Missing User Warnings

Low

Confidence: 78% confidence
Finding: The example saves a local .docx file but provides no warning about overwriting existing files or modifying user data. While the example is simple, omission of file-safety guidance in a write-capable skill can still cause accidental data loss.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.