colima-prompt-handoffs procedure — revision 10f80d6e

Procedure colima-prompt-handoffs · revision 10f80d6e · 2026-09-13

What this test proves

Inside the shared colima testbed, a real agent executes prompt AT11 and observes only public ATM CLI JSON. Task events expose queued, ready, reminder, and started for the prompted task but queued only for later tasks; disabling reminders removes the reminder event and produces the documented doctor finding.

Evidence layout

This procedure is one step of a colima integration run. The agent's unedited prompt-report-1 JSON is retained beside step.json; scripts/integration/render_colima.py renders the panel and aggregate. Historical payloads remain byte-identical and render through the same path.

Changes

The revision sections below list every runner revision that produced committed evidence, newest first. A run whose source revision is not an ancestor of any listed revision links to the revision in effect on its run date and is marked inferred.

Flow

Reuse the live testbed fixture

Run AT11 prompt-handoffs agent prompt

1: prompted task events only

2: disabled reminder doctor finding

Aggregate PASS/FAIL

Write prompt report and step.json

runner revision 10f80d6e

Steps

stepactionobservableevidence
1Tell the fixture agent to assign three tasks, wait for a reminder, and start the prompted task`atm task events --json` shows queued, ready, reminder, started for task 1 and queued only for tasks 2–3`prompt-AT11.json` and `step.json` cases
2Tell the agent to disable task reminders and assign another tasktask events omit reminder and `atm doctor --json` reports `disabled_task_nudge_template_override``prompt-AT11.json` and `step.json` cases

Revision note: refactor(bb8): drive colima checks through fixture prompts