Security audit

命令行工具

Security checks across malware telemetry and agentic risk

Overview

This skill is a Google Workspace command guide, but it gives broad agent access to read and change mail, files, contacts, calendars, spreadsheets, and local outputs without consistently tight confirmation rules.

Install only if you want an agent to use the gog CLI with your Google Workspace account. Authorize the minimum Google services needed, avoid --no-input for sends/writes/clears/exports unless you have reviewed the exact command, and require the agent to state the account, service, target file or sheet, and side effects before it runs anything.

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: 94% confidence
Finding: The skill claims to be pure Markdown/natural-language driven while the metadata and body also expose `exec`-style command execution and shell workflows. This inconsistency can mislead an agent or reviewer into applying a weaker trust model, increasing the chance that command execution is enabled without appropriate safeguards.

Intent-Code Divergence

Medium

Confidence: 89% confidence
Finding: The document states Calendar only supports querying events, but elsewhere references creating events and needing confirmation before doing so. Contradictory action boundaries create operator confusion and can cause an agent to perform state-changing actions under the false assumption that the skill is read-only for that service.

Vague Triggers

High

Confidence: 96% confidence
Finding: The activation criteria are extremely broad, effectively saying to use the skill for general productivity, automation, batch processing, or workflow optimization. Because the skill has access to external Google Workspace actions and command execution, vague triggering increases the risk of the agent invoking it in contexts the user did not specifically authorize, including data access and modifications.

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The operating instructions tell the agent to execute tasks based on general user descriptions rather than explicit confirmation of target service, account, and side effects. In a skill that can search mail, read documents, and send mail, this increases the chance of overbroad or unintended tool use.

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: The skill spans Gmail, Calendar, Drive, Contacts, Sheets, and Docs, including write-capable operations like sending mail and updating spreadsheets, but the description does not prominently warn that it can access, transmit, or alter user data in third-party services. Users may therefore authorize or invoke it without understanding the privacy and integrity risks.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.