Lesson 3 4 min
Try it and read the work
Try it runs the real agent. Learn to read what it shows you, and what its eyes are saying.
The real agent
Try it runs exactly what a chat with the agent would use: its model, skills, rules, approvals and max steps. A row of chips shows what is in effect: the model and where it runs, each skill (marked off when it can’t be used), the rules preset and the step limit. Reset starts over with a fresh chat, tools and browser session.
Reading the work
You see the same work timeline as in a chat: each tool step as it happens, approval cards with their receipts, and a banner when a skill is off. When a reply uses all its steps, a Not done yet card offers Continue, Finish with what you have and Change the limit.
What the eyes say
Every agent is a character, and its eyes show what it is doing wherever it appears.
| Eyes | Meaning |
|---|---|
| Blink and glance around | Idle |
| Drift upward | Thinking |
| Scan side to side | Working: browsing or using a tool |
| Look at you and bounce | Needs your ok |
| Happy squint and a hop | Done |
| Worried brows | Can’t: a step failed or a skill is unavailable |
| Closed | Offline: its model isn’t ready |
To calm the motion, go to Settings › Preferences › Agent animations and choose Lively, Calm or Off.
Key takeaways
- Try it uses the agent's real model, skills, rules, approvals and step limit, so actions are real.
- Chips show what is in effect; a banner says when a skill is off.
- The eyes show state everywhere; looking at you and bouncing means it needs your ok.