Agent Journey Lab · Journey run record
Completed example record
This page shows one complete record from the calibration fixture in our own repository. It shows a recorded stop, and it shows the clean completion state. It is internal fixture evidence. It is not a customer result and it is not a provider-agent result.
Internal fixture record — not a customer result.
Start a $199 pilotRead the pilot scope
Our test sessions declare AgentJourneyLab/0.1. Site owners can stop a public test: read the research boundary or write to research@agentjourneylab.com.
What this is
What this record is
Internal fixture record — not a customer result.
Every value on this page comes from the calibration fixture in our own repository. The fixture is a local test site on a loopback address. Our deterministic test script made the attempts in a real browser engine. No AI model made these attempts, and no provider agent made them. No customer site was touched.
What the fixture produced: the goal, the checkpoint states, the observation, the control result, the configuration, the build identity, and the completion evidence. What is illustrative: the n of 3 shape for this second journey, because the committed fixture runs it one time. A customer baseline runs the primary configuration three times.
Header fields
Header fields
Internal fixture record — not a customer result.
| Record ID | AJL-FIXTURE-CAL-002 |
|---|---|
| Journey ID | ambiguous_label_records_friction |
| Goal | Renew a plan and reach the page that carries the reference AJL-FIXTURE-OK. |
| Target | Local fixture site on a loopback address. No public domain was used. |
| Attempts | 1 attempt. The committed fixture runs this journey one time. |
| Configuration | Primary configuration — the deterministic fixture script in a real browser engine. |
| Runner version | runner-v1.0.0 |
| Prompt version | prompt-v1 |
| Build identity | calibration-runner-v1 |
| Environment fields | A record carries the browser version, the viewport, the locale, the timezone, the start time in UTC, and the network vantage. This page prints no value for them, because the committed fixture fixes no value for them. |
| Control journey | Passed. The clear fixture journey completed in 3 of 3 attempts in the same batch, so the harness was working when this stop was recorded. |
| Operator | [redacted] |
The token [redacted] is the redaction token that our evidence code writes in place of any value that must not leave the browser. It is the only token of that form on this page.
Checkpoints
Checkpoints
Internal fixture record — not a customer result.
- 01Entry page✓ reached
- 02Renew a plan✗ stopped
- 03Confirmation page· not reached
The attempt reached the entry page. It stopped at the second checkpoint. It never reached the confirmation page.
Observation
Observation at the stop
The page /ambiguous/ offers two controls with the same visible label, Continue. Both controls point at the same page. The agent followed one control and the page did not change. The reference AJL-FIXTURE-OK was not on the page, and the page states in words that the reference is not there.
Final state: the entry page of the ambiguous journey, after one control was followed. The declared completion evidence was absent.
Outcome and control
Outcome label and control result
Internal fixture record — not a customer result.
| Outcome label | Reproducible agent friction — candidate. One attempt is not a repeat, so this label stays a candidate until three attempts agree. |
|---|---|
| Control result | Passed. The clear fixture journey completed in 3 of 3 attempts in the same batch. |
| Confidence | Medium. The observation is clear, and the attempt count is one. |
| Change since last run | First record for this journey. There is no earlier record to compare. |
Likely fix owner
Likely fix owner
Front end — the two controls that carry the same label and the same destination.
Verification step: give the page to the team that owns the template, and ask for one control per destination with a different label. A likely owner is not a verdict, and this record includes no remediation work.
Self-check
Self-check
These steps repeat the stop in less than one minute.
- Open the fixture page for the ambiguous journey.
- Read the two links. Both carry the label Continue.
- Follow one link. The page does not change.
- Search the page for the reference AJL-FIXTURE-OK. It is not there.
Limits
Limits of this record
This record covers one local fixture journey, one configuration, one attempt, and one batch. It states nothing about a public website, nothing about a provider agent, and nothing about every agent. It is not a certificate.
Clean completion
Clean completion state
This is what you receive when the journey completes and no repeated stop occurs. The same fixture batch produced it.
Internal fixture record — not a customer result.
| Record ID | AJL-FIXTURE-CAL-001 |
|---|---|
| Journey ID | clear_journey_completes |
| Goal | Book a demo slot and reach the confirmed booking page. |
| Result | Completed in 3 of 3 attempts. |
| Outcome label | Consistent completion. |
| Completion evidence | The page /clear/done.html carries one visible element with the approved state key confirmed and the exact reference AJL-FIXTURE-OK. |
| Tested conditions | Primary configuration, runner-v1.0.0, prompt-v1, build calibration-runner-v1, three fresh browser contexts, one batch. |
| Limits | This result covers one local fixture journey, one configuration, and one batch. It states nothing about a public website and nothing about every agent. |
| Reference state for monitoring | Four checkpoints reached, and the completion evidence present. A later weekly record is compared with this state. |
Internal fixture record — not a customer result.
- 01Entry page✓ reached
- 02Choose the slot✓ reached
- 03Confirm the slot✓ reached
- 04Booking confirmed✓ reached
Four checkpoints reached, in each of the three attempts.
How to read a real record
How to read a real record
A customer record carries the same fields, with the customer journey, the named provider and model alias, the dated attempts, and the environment fields. It reports completion as n of 3 for the primary configuration, and it reports one breadth attempt for the secondary configuration. A record covers only the configurations and the dates that it names.
To keep this record, print this page. The page needs no script and no third-party file, and it prints on one document register.
Next step