Security audit

Windows应用控制器

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed Windows desktop automation guide, but it gives an agent broad click, typing, screenshot, file-saving, and force-close abilities without enough consent or scoping safeguards.

Install only if you intentionally want an agent to control a Windows desktop. Use it in a VM or test account when possible, close sensitive windows first, and require explicit confirmation before screenshots, clipboard use, form submission, file creation or overwrite, app launch, or force-closing processes.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (4)

Vague Triggers

Medium

Confidence: 84% confidence
Finding: The activation phrases like “控制应用 / 自动化Windows / GUI操作” are broad enough that ordinary user requests about Windows usage could unintentionally trigger a high-privilege desktop automation skill. Because this skill can launch/kill processes, type keystrokes, click UI elements, and capture screenshots, accidental invocation materially raises the risk of unintended destructive or privacy-impacting actions.

Missing User Warnings

High

Confidence: 95% confidence
Finding: The skill advertises capabilities to open/close apps, click through GUIs, fill forms, and take screenshots, but it does not provide prominent user-facing warnings or consent requirements for destructive operations and privacy-sensitive data capture. In the context of desktop automation, these capabilities can expose on-screen secrets, alter user data, submit forms, or terminate applications without sufficiently informed approval.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The example workflow saves a file and screenshots to disk without warning that it will create or overwrite local artifacts. Even as sample code, this normalizes state-changing behavior in a skill intended for autonomous use, increasing the chance of unintended data modification, persistence of sensitive screenshots, or cluttering user systems.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The screenshot examples include full-screen and region capture, plus saving outputs, without privacy warnings or consent checks. In a desktop-control skill, screenshots can capture credentials, personal messages, financial data, or proprietary information from any visible application, making this especially dangerous in unattended or ambiguous contexts.

Static analysis

No suspicious patterns detected.