The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill metadata specifies an installation step that clones the core ATL repository from the author's GitHub account (https://github.com/JordanCoin/Atl).
[COMMAND_EXECUTION]: During setup, the skill executes a shell script (scripts/setup.sh) that uses 'xcodebuild' to compile the automation server and 'xcrun simctl' to manage the iOS Simulator environment and install the resulting app.
[COMMAND_EXECUTION]: The skill uses 'curl' to communicate with local HTTP servers running on ports 9222 and 9223 to send automation commands (navigation, clicks, screenshots) to the simulator.
[PROMPT_INJECTION]: As a browser and app automation tool, the skill is inherently exposed to Indirect Prompt Injection. It ingests untrusted data from web pages and accessibility trees which could contain instructions intended to manipulate the agent's behavior. The skill includes some guidance for agents to verify states, which can mitigate accidental obedience but does not fully prevent adversarial injection.

atl-browser