WEBVTT

NOTE Edited evidence walkthrough using recorded native-client reports and an actual reviewer-inbox capture. Synthetic data only. The separate reviewer was an automated test reviewer; this is not a continuous screen recording.

1
00:00:00.000 --> 00:00:05.000
Edited evidence walkthrough.
Synthetic data; automated test reviewer.

2
00:00:05.000 --> 00:00:12.000
A real Claude Code client reads the bounded chart
and submits a draft through Last's MCP connector.

3
00:00:12.000 --> 00:00:20.000
The client exits while the draft awaits review.
No FHIR Task has been saved by this work item.

4
00:00:20.000 --> 00:00:27.000
Last's separate reviewer inbox shows the exact draft,
source version, and proposal hash.

5
00:00:27.000 --> 00:00:34.000
An automated test reviewer approves the draft.
The assistant has no reviewer key or approval tool.

6
00:00:34.000 --> 00:00:40.000
Last independently verifies Task/1034, version 1.
The receipt records a requested follow-up.

7
00:00:40.000 --> 00:00:48.000
A fresh Claude Code process reconnects using get_work
and retrieves the result for the same Work ID.

8
00:00:48.000 --> 00:00:55.000
An independent FHIR GET finds exactly one matching Task
and confirms its fields match the approved draft.

9
00:00:55.000 --> 00:01:00.000
The assistant reports the verified save.
Patient contact has not been performed.