Security audit

屏幕监控工具-免费版

Security checks for vulnerabilities and agentic risk

Overview

This skill is a screen-capture and timed-monitoring helper, but its activation rules and privacy/data-flow disclosures are too broad for that level of access.

Review before installing. Only use this skill when you intentionally want the agent to capture your screen, avoid sensitive windows, set short durations and explicit output directories, delete saved screenshots when done, and clarify whether any analysis could send screenshot content to an LLM or external service.

SkillSpector

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 (4)

Intent-Code Divergence

Medium

Confidence: 91% confidence
Finding: The documentation is internally inconsistent: one section says the skill does not use external APIs, while another lists an LLM API as required. This can mislead users and reviewers about data flow and trust boundaries, especially for a screen-capture skill that may expose sensitive on-screen content to an LLM or agent runtime.

Vague Triggers

High

Confidence: 95% confidence
Finding: The trigger condition is extremely broad ('需要项目管理、任务规划、进度跟踪、团队协作时使用'), which could cause the agent to invoke a screen-monitoring skill in many unrelated situations. For a capability that captures screens and performs scheduled monitoring, ambiguous activation materially increases the risk of unintended surveillance, privacy violations, and excessive data collection.

Vague Triggers

Medium

Confidence: 92% confidence
Finding: Stating that users can trigger the scenario through natural-language instructions without defining concrete phrases or boundaries makes invocation too permissive. In the context of a screen-capture tool, this can lead the agent to infer consent from vague requests and perform captures when the user did not clearly authorize them.

Missing User Warnings

High

Confidence: 97% confidence
Finding: The skill advertises screenshot capture, timed monitoring, local saving, and content analysis without an upfront warning about privacy risks, retention, and possible sensitive-data collection. Because screenshots can include credentials, personal data, internal documents, or regulated information, missing notice and consent controls is dangerous in this context.

Static analysis

No suspicious patterns detected.