Recording a demo

A skill is only as good as the demonstration behind it. This page covers how to record one cleanly.

Recording a demonstration

1. Start (or join) a session and open the app you want to automate.

2. Begin recording. Tell the agent in chat, e.g. record a demo, or press the record button (⏺) on the stream. The badge turns red while capturing.

3. Do the task once, deliberately. Tap, type, and scroll the way you want the skill to behave. Move at a normal pace — there’s no time pressure.

4. Stop recording. Say stop recording or press the button again. The induction pipeline runs and proposes a skill.

What gets captured

• Touch events (getevent) — taps, swipes, and their coordinates
• Screenshots at each step, for selector inference
• App + window state — package and activity at every transition
• Typed text — including anything you enter into fields

Tips for a skill that generalizes

• One task per recording. Don’t chain unrelated flows; record them separately so each becomes its own reusable skill.
• Start from a stable screen (e.g. the app’s home), so replay has a predictable entry point.
• Prefer visible, labeled controls over tiny or ambiguous tap targets — they produce more robust selectors.
• Parameterize after the fact. Record with concrete values; the pipeline detects which inputs are variable so the skill can be reused with new ones.

Recordings can contain secrets

A recording captures everything you type, including passwords or codes entered during the demo. Recordings are never committed and, in the hosted product, are only uploaded if you explicitly opt in. See Privacy & data.

Next: Induction pipeline