The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The Python script scripts/xcode.py downloads and executes the mcporter@0.13.11 package from the public npm registry at runtime using the npm exec command.
[COMMAND_EXECUTION]: The skill uses subprocess.run within scripts/xcode.py to execute shell commands, including the npm binary and the Xcode bridge service xcrun mcpbridge (as configured in mcp.json).
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection through the processing of Apple-platform project data and external tool arguments.
Ingestion points: External data is ingested from a JSON file specified by the --args-file parameter in scripts/xcode.py and from the output of Xcode tools.
Boundary markers: No explicit delimiters or instructions are used to protect the agent from malicious instructions embedded within the processed project files or tool arguments.
Capability inventory: The skill can execute local commands via subprocesses and perform file system operations through the Xcode toolchain.
Sanitization: Input from the arguments file is validated for valid JSON structure, but the content is not sanitized for potential injection patterns before being passed to the toolchain.

xcode