Security audit

办公效率中枢

Security checks across malware telemetry and agentic risk

Overview

This office automation skill is mostly purpose-aligned, but it needs review because it combines file access, command execution, email/cloud/database integrations, and scheduled automation without tight scoping or mandatory confirmation steps.

Review before installing. Use this only where the agent is allowed to process office files and sensitive business data, and require dry runs, recipient review, attachment/content previews, least-privilege SMTP/database/cloud credentials, and explicit approval before emails, uploads, database queries, or scheduled cron jobs run.

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 (6)

Intent-Code Divergence

Medium

Confidence: 91% confidence
Finding: The skill states that data is processed locally by default, but other sections explicitly enable SMTP, webhooks, cloud storage, database access, and automated report distribution. This mismatch can mislead users into sharing sensitive office documents or PII under a false assumption that no external transmission will occur.

Intent-Code Divergence

Medium

Confidence: 95% confidence
Finding: The file contradicts itself by first declaring the skill as MD+EXEC requiring command/script execution, then later describing it as pure Markdown only. This ambiguity is dangerous because users or calling agents may underestimate the skill's ability to execute commands, leading to unsafe invocation or overly broad trust.

Vague Triggers

Medium

Confidence: 87% confidence
Finding: The natural-language invocation examples are broad enough to match ordinary office requests such as document conversion, emailing, scheduling, and reporting. In an agent environment, this can cause overbroad or accidental activation, especially because the skill has write and exec capabilities and may handle sensitive business data.

Vague Triggers

Medium

Confidence: 89% confidence
Finding: The description says to use the skill whenever efficiency, automation, batch processing, or workflow optimization is needed, which is far too broad for a powerful office automation skill. This increases the chance that benign everyday requests get routed into a capability set that can read/write files, execute commands, and send external communications.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The skill describes automated email sending and report distribution but does not provide a prominent warning that these actions may transmit potentially sensitive document contents or attachments to external recipients. In office workflows, this creates a realistic risk of unintended data disclosure, especially when batch processing and templated distribution are involved.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The workflow combines database collection, automated report generation, and emailing to recipients without a strong privacy warning or recipient-verification control. This is particularly dangerous in the office context because internal databases may contain sensitive business or personal data, and a wrong recipient list or automatic send could cause a serious confidentiality breach.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.