The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a dedicated CLI wrapper (orca) to execute simulator commands. This includes high-level interactions like tap, type, and button, as well as a raw execution mode (exec --command) for advanced SimulatorKit features.
[EXTERNAL_DOWNLOADS]: References the serve-sim open-source tool and its GitHub repository (github.com/EvanBacon/serve-sim). The skill mentions npx serve-sim as a fallback, which is a standard method for executing Node.js utilities from the official registry.
[DATA_EXFILTRATION]: Contains instructions for camera injection using local files (--file /tmp/test.mp4). This is a legitimate feature for mobile testing and does not involve exfiltration of sensitive user data.
[PROMPT_INJECTION]: The ax command extracts the accessibility tree from running applications. This represents an indirect prompt injection surface as the agent ingests external, untrusted UI data.
Ingestion points: UI hierarchy data ingested via ORCA emulator ax (SKILL.md).
Boundary markers: Absent.
Capability inventory: tap, gesture, type, and exec commands (SKILL.md).
Sanitization: Not specified.

orca-emulator