The Agent Skills Directory

[COMMAND_EXECUTION]: The skill orchestrates the execution of system-level Xcode and simulator commands via MCP tools such as build_ios_sim_app, boot_simulator, and install_app_on_simulator. These operations are essential for its primary function of mobile app testing.
[EXTERNAL_DOWNLOADS]: The workflow documentation recommends installing the xcodebuildmcp tool using Homebrew (getsentry/xcodebuildmcp) or npx (xcodebuildmcp@latest). These resources are hosted by Sentry, a reputable service provider, and follow standard software installation practices for developers.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes application logs (get_sim_logs) and system output without explicit sanitization or strict boundary markers. An application being tested could theoretically output malicious instructions that an agent might inadvertently follow.
Ingestion points: Raw application and simulator logs retrieved through the get_sim_logs tool referenced in SKILL.md.
Boundary markers: None present; the agent is instructed to "Check logs for errors" without delimiters or instructions to ignore embedded commands.
Capability inventory: The agent has access to a broad set of developer tools via the XcodeBuildMCP server, including building apps, managing simulators, and capturing logs, as well as file system discovery tools.
Sanitization: No filtering or validation logic is applied to the log content before it is parsed by the agent.

test-xcode