The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's primary functionality is built upon the execution of system-level commands to control the iOS simulator environment. It uses the subprocess module to invoke xcrun simctl for app management and device lifecycle, idb (iOS Development Bridge) for accessibility-driven UI interactions, and xcodebuild for project compilation and testing. These operations are essential to the skill's purpose and are implemented using best practices (passing commands as lists rather than shell strings).
[INDIRECT_PROMPT_INJECTION]: Several scripts ingest data from the simulator that is technically outside the skill's direct control. Specifically, log_monitor.py and app_state_capture.py stream or capture system and app logs, while navigator.py and screen_mapper.py parse UI accessibility trees. If a monitored application generates malicious text in its logs or accessibility labels, it could potentially influence the behavior of the AI agent interpreting these outputs.
Ingestion points: scripts/log_monitor.py (simctl log stream), scripts/app_state_capture.py (simctl log show), scripts/navigator.py and scripts/common/idb_utils.py (idb ui describe-all).
Boundary markers: None. Data extracted from logs and UI hierarchies is presented to the agent context without specific encapsulation or sanitization.
Capability inventory: The skill possesses extensive capabilities including file system access for saving artifacts, and the ability to execute arbitrary iOS simulator commands (tap, type, launch, etc.).
Sanitization: No significant sanitization or filtering is applied to the content of logs or UI labels before they are returned to the agent context.

automating-ios-simulator