Back to the run
Unexpected modal · run 2
Failedunexpected_uiDEMO DATAAn overlay covered the page and 3 clicks never reached the intended element.
Duration
28.4s
Steps
2
Evaluator score
30%
partial credit
Session
—
seeded demo — no session
Session replay
A DOM-level rrweb recording of exactly what the browser did.
Seeded demo runs carry no replay.
Replays come from real browser sessions. Start a run to record one.
Evaluator evidence
Read from the benchmark site's server-side state, independently of the agent.
| Assertion | Expected | Actual | Result |
|---|---|---|---|
aurora-headphones is in the cart product_in_cart | true | false | fail |
cart holds 1 of aurora-headphones quantity | 1 | 0 | fail |
coupon SAVE20 is applied coupon_applied | SAVE20 | null | fail |
the discount is reflected in the total discount_applied | true | false | fail |
checkout name is "Ada Lovelace" checkout_name | Ada Lovelace | null | fail |
checkout city is "London" checkout_city | London | null | fail |
the flow reached the "review" stage reached_stage | review | browse | fail |
the order was NOT placed purchase_not_submitted | false | false | pass |
What the agent said about itself
max_steps — “Used all steps without reporting completion.”
Recorded as a claim. The verdict above comes from the site's own state.
Agent action trace
2 actions
- 1click e9 succeededclicked button "Accept all"
- 2click Add to cart: Aurora Headphones failedno element matching "Add to cart: Aurora Headphones"
Timeline
Every lifecycle beat, in order.
- 04:56.863Lifecyclerun_queued
- 04:56.983Lifecycleenvironment_preparing
- 04:57.836Lifecyclefixture_ready
- 04:59.126Lifecyclesession_created
- 04:59.426Navigation/?run=demo
- 04:59.476Lifecycleagent_started
- 05:01.543Lifecycleagent_finished
- 05:01.663Lifecycleevaluation_started
- 05:02.038Evaluator{"score":0.5,"success":false}
- 05:02.058Lifecycleevaluation_finished
- 05:02.358Lifecyclebrowser_released
- 05:02.398Lifecyclecleanup_complete
Console and page errors
- errorUncaught (in promise) intercepted pointer event
Failed requests
No request failures.