Security audit

投屏控制免费版

Security checks for vulnerabilities and agentic risk

Overview

This Chromecast control skill appears non-malicious, but its broad routing instructions, inconsistent security claims, and local-network/file-casting authority need human review before installation.

Install only if you intend to use it for Chromecast control with catt on a trusted local network. Avoid casting sensitive local files or internal URLs, review commands before execution, and do not rely on the API-key or HTTPS safety claims as written.

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 (7)

Intent-Code Divergence

Medium

Confidence: 95% confidence
Finding: The document gives contradictory security and dependency claims: one section says no API key is needed and all functionality is local, while other sections instruct users to configure API keys and call API services. This inconsistency can mislead users and agents about trust boundaries, secrets handling, and whether networked services are involved, increasing the chance of unsafe deployment or accidental credential disclosure.

Intent-Code Divergence

Medium

Confidence: 97% confidence
Finding: The skill claims command execution will avoid concatenating user input, but its examples and behavior require passing user-controlled URLs, file paths, and device names into shell commands. In an MD+EXEC skill with Bash access, that mismatch is dangerous because an agent may naively interpolate untrusted input into a shell invocation, creating command injection risk or unintended local file/network access.

Intent-Code Divergence

Medium

Confidence: 89% confidence
Finding: The document states network communication is forced over HTTPS with SSL verification, but the actual skill supports local-network discovery, local file casting, and arbitrary URLs, including non-HTTPS scenarios. This false security claim can cause users to overtrust the confidentiality and integrity of content transmission, especially on shared or insecure LANs.

Description-Behavior Mismatch

High

Confidence: 91% confidence
Finding: The startup trigger says the skill should be used for database operations and SQL tasks, which is unrelated to Chromecast control. In an agent ecosystem, incorrect activation conditions can cause the skill to be invoked in the wrong contexts, leading to unintended command execution, device scanning, or exposure of unrelated user data to a tool with Bash privileges.

Vague Triggers

High

Confidence: 93% confidence
Finding: The activation description is overly broad, framed around general efficiency, automation, and workflow optimization rather than specifically Chromecast control. Because the skill has Bash capability, broad triggering increases the risk that an agent selects it for unrelated tasks and executes commands or performs network actions outside user intent.

Vague Triggers

High

Confidence: 95% confidence
Finding: The startup trigger section explicitly references unrelated database and SQL use, making skill routing ambiguous and unsafe. In context, this is more dangerous because the skill can execute Bash and interact with the local network, so accidental invocation can lead to unintended scans, casts, or command execution in response to non-Chromecast requests.

Missing User Warnings

Medium

Confidence: 84% confidence
Finding: The skill does not clearly warn that casting local files or URLs transmits content over the local network to Chromecast devices. This omission matters in a media-casting tool because users may unintentionally expose sensitive local files, internal URLs, or viewing activity to devices and network observers on the same LAN.

Static analysis

No suspicious patterns detected.