ios-simulator-browser

Warn

Audited by Gen Agent Trust Hub on Nov 4, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructions in SKILL.md specify the use of npx --yes serve-sim@latest, which fetches and executes a package from the npm registry at runtime. Running unversioned packages from remote registries introduces a supply chain risk.
  • [REMOTE_CODE_EXECUTION]: The use of npx to run an external utility constitutes remote code execution, as the package content is retrieved from a remote server and executed on the local machine.
  • [COMMAND_EXECUTION]: The scripts/swiftui-preview-browser.mjs script makes extensive use of system commands to interact with the iOS Simulator and build tools:
  • It uses xcrun simctl for simulator lifecycle management, including boot, install, launch, terminate, and get_app_container.
  • It uses xcodebuild to compile both the generated host application and the hot-reload dynamic library.
  • It uses swift package commands (dump-package, describe) to extract metadata from the user's project.
  • It uses notifyutil to trigger hot-reload events within the simulator process.
  • [DYNAMIC_EXECUTION]: The skill performs complex runtime code generation and compilation:
  • It generates Swift source code for a host application and a reload plugin dylib based on the target package's metadata.
  • It dynamically constructs an Xcode project file (project.pbxproj) to coordinate the build.
  • It compiles these generated artifacts at runtime using xcodebuild and injects the resulting binaries into the simulator process.
  • It relies on a local module ./lib/xcode-project.mjs (not included in the analyzed files) to handle project generation logic.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process user-provided Swift packages, which creates an attack surface for malicious code to influence the build or execution environment.
  • Ingestion points: The script reads and parses Package.swift and scans the package directory for source files.
  • Boundary markers: No boundary markers or "ignore instructions" warnings are utilized when processing content from the Swift package.
  • Capability inventory: The script has broad capabilities including file system access, network-adjacent operations (via npx), and full command execution for building and running code.
  • Sanitization: Metadata extracted from the package, such as target names, is interpolated into code templates and command-line arguments without explicit sanitization.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Nov 4, 2026, 10:12 AM