building-apple-platform-products

Warn

Audited by Gen Agent Trust Hub on Oct 1, 2026

Risk Level: MEDIUMPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PRIVILEGE_ESCALATION]: The skill provides instructions for the agent to perform administrative tasks using sudo. Specifically, it includes commands for switching the system-wide active Xcode version using sudo xcode-select -s and accepting the Xcode license with sudo xcodebuild -license accept (SKILL.md, references/ci-cd-patterns.md, references/xcrun-basics.md, references/troubleshooting.md). While these are standard development tasks, they involve modification of system-level configurations.
  • [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion and processing of metadata from local project files, which represents a potential injection vector if project or target names are maliciously crafted. * Ingestion points: Discovery commands such as xcodebuild -list, swift package describe, and ls (SKILL.md, references/project-discovery.md). * Boundary markers: No delimiters or specific safety warnings are provided for the output of these discovery commands. * Capability inventory: The skill has extensive shell execution capabilities including build tools (xcodebuild, swift), simulator management (simctl), device deployment (devicectl), and security utilities (security, codesign). * Sanitization: No sanitization or validation logic is defined for the project structure data retrieved during discovery.
  • [COMMAND_EXECUTION]: The skill facilitates the execution of powerful system utilities for automation. This includes security for keychain creation and certificate importation in CI environments, xcrun devicectl for file transfers to and from physical devices, and simctl spawn for executing arbitrary processes within a simulator (references/code-signing.md, references/devicectl.md, references/simctl.md).
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 1, 2026, 08:49 PM