Security audit

Excel Craft Free

Security checks across malware telemetry and agentic risk

Overview

This skill is a local Excel workbook generator with some overly broad wording, but no evidence of hidden data access, persistence, exfiltration, or destructive behavior.

Install only if you want an agent to generate and run local Python/openpyxl scripts for Excel files. Use it for explicit workbook-generation tasks, specify the output path, review generated code before execution, and avoid relying on it for general document conversion or content extraction despite the broad wording.

SkillSpector

By NVIDIA

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
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 (5)

Intent-Code Divergence

Medium

Confidence: 90% confidence
Finding: The skill states that it does not involve network calls or external endpoints, but later instructs the agent/user to run ping for troubleshooting. That inconsistency can mislead users and policy engines about the skill’s actual operational scope, and may cause an agent with exec capability to perform unexpected network-related actions. In a skill that already has exec access, even small scope drift matters because troubleshooting steps can normalize broader command execution than the declared purpose requires.

Description-Behavior Mismatch

Medium

Confidence: 92% confidence
Finding: The manifest advertises an Excel-generation skill, but the description expands its trigger scope to generic file processing, document conversion, format conversion, and content extraction. This overbroad framing can cause an agent to invoke the skill in contexts beyond its documented implementation, increasing the chance of inappropriate file access, misuse of exec, or user confusion about what actions the skill may take. Because the skill includes read/write/exec tools, vague scope inflation is more dangerous than it would be in a read-only skill.

Vague Triggers

High

Confidence: 95% confidence
Finding: The skill description uses an overly broad trigger scope covering generic file-processing tasks, which can cause the agent to select this skill for requests far outside Excel generation. Overbroad activation is dangerous because the skill has read, write, and exec permissions, so mistaken invocation can lead to unnecessary local file creation, access, or code execution in unrelated workflows. In this context, the mismatch between narrow implementation and broad routing language materially increases operational risk.

Vague Triggers

Medium

Confidence: 88% confidence
Finding: The capability coverage section lists a large set of vague keywords instead of clear activation conditions. This can bias a planner or router toward selecting the skill on loosely related prompts, especially where terms like '文件处理' or '内容提取' appear, despite the actual skill being focused on spreadsheet generation. Given the available exec/write tools, ambiguous activation criteria increase the risk of unintended use and unnecessary file operations.

Missing User Warnings

Medium

Confidence: 84% confidence
Finding: The markdown does not clearly warn users that the skill writes and saves local Excel files. Without an upfront notice, users may not understand that invoking the skill can create or overwrite artifacts on disk, which is an important safety boundary for a tool with write and exec capabilities. In this context, omission of side-effect disclosure increases the risk of unintended local modifications.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.