Describe one small game or training interaction to Codex or Claude Code and supply the documented interface. Start with explicitly simulated values, then verify with real hardware. AI-generated code still needs installation, integration and testing; do not invent physiological thresholds or clinical outcomes.
A compatible BioSignals application runs locally and can expose a WebSocket service at ws://localhost:8080. Use the path, channels, JSON fields, units and access controls documented for that exact application/version. Observe connection state, missing values and signal quality; do not assume every sensor or camera field is available.
Supported applications can combine sensor signals with visible eye, mouth and head movements from a standard camera, with user permission. Facial motion is contextual information, not certain emotion recognition. Do not infer a camera schema from sensor messages. Keep physiological data local unless there is an authorized, documented sharing arrangement.
Check app startup, connection, incoming messages, units, timestamps, reconnection and disconnection behavior. Validate the experience first with clearly marked simulation and then with the actual sensor. Record the tested app version and hardware. Confirm browser/platform restrictions before relying on a local connection from a hosted HTTPS page.