swift-macos

Security checks across malware telemetry and agentic risk

Overview

This is a documentation-only Swift/macOS development skill that covers sensitive APIs, but the behavior is disclosed and aligned with its stated purpose.

Install/use this skill only if you want a broad Swift/macOS development reference. Review generated code and commands before running them, especially anything that records screen/audio, requests microphone or TCC permissions, registers login items or agents, creates daemons, stores Apple credentials, signs/notarizes apps, or executes remote installer commands.

SkillSpector

By NVIDIA

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (5)

Context-Inappropriate Capability

Medium

Confidence: 84% confidence
Finding: The documentation goes beyond ordinary macOS app integration and includes concrete guidance for LaunchAgents and LaunchDaemons, including root-running services. While these APIs are legitimate, documenting them without strong scope limits, trust boundaries, and user-consent guidance increases the chance that downstream agents or developers will build persistent privileged components that are unnecessary for the stated skill scope.

Missing User Warnings

Medium

Confidence: 84% confidence
Finding: The skill provides ready-to-use ScreenCaptureKit examples for screen, app-audio, and microphone capture without pairing them with explicit privacy, consent, and TCC-permission guidance. In a code-generation context, this can normalize or accelerate implementation of surveillance-capable features that capture sensitive user data, especially because the examples include audio capture and target-app filtering.

Missing User Warnings

Medium

Confidence: 98% confidence
Finding: The document recommends `curl -L https://swift.org/install | bash`, which executes remotely fetched content immediately without verification, pinning, or inspection. Even though this appears intended as convenience documentation, it normalizes a risky installation pattern that could lead to arbitrary code execution if the remote endpoint, transport chain, or distribution process is compromised.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: The process-observation and per-process audio examples describe how to enumerate running apps and infer which apps are actively using audio I/O, which can reveal sensitive user behavior such as calls or app usage patterns. The lack of an explicit privacy notice, consent flow, retention limits, and platform-policy warnings makes this dangerous because it normalizes surveillance-capable features without safeguards.

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: The login item, agent, daemon, and background-app sections provide ready-to-use persistence mechanisms but do not clearly warn that these create continuous or login-triggered background execution on the user's system. In practice, this lowers the barrier to implementing stealthy persistence patterns, especially when combined with LSUIElement or helper/agent deployment.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal