Mark J. SalisburyMJS-OM-2026 · Sheet 02
Just show me the cool stuff
Teaching · A working template

A build record you can reuse.

Keep the job, decisions, checks and handoff together. Leave unknowns visible.

Use this for a small tool, a workflow or a release. Replace the prompts with evidence from your own work. An unknown date or an unrun check should stay unknown or unrun.

The record can grow as the work develops. It should make the next decision easier: release, repair, investigate, pause, or ask the person who will use it.

The eight parts

  1. The job and constraints. Who needs it, what they need to do, what is out of scope, and what must not change.
  2. The version and dates. Identify the build. Keep the original publication date, meaningful record-update date, and actual check dates separate.
  3. The decisions. What was chosen, which useful alternatives were set aside, and why. Include a change of direction when it explains the result.
  4. The checks. Record expected and observed behavior, the command or procedure, the build and device or environment, and the evidence. State the scope of an independent review.
  5. The failures. Explain what failed, how it was detected, what changed, and what was checked after the fix. Keep an assisted repair distinguishable from the first attempt.
  6. The limits and human acceptance. What remains untested? What would block release? Who must make the remaining decision?
  7. The operation and handoff. How the owner checks health, recovers from a failure, and knows when to stop. Keep authority and recovery steps explicit.
  8. The evidence and publication boundaries. Link supporting records, credit collaborators and assets, and identify what may be shared. Keep private data and credentials out of the public record.

A fictional example, before testing

Fictional planning example

The room-booking checklist below is invented for this template. Its inputs and expected behavior are fictional. No build, test, customer outcome or release is being reported.

Job: Flag overlapping bookings in the same room before a coordinator confirms the schedule. Use fictional booking IDs only. The checklist cannot send or confirm a reservation.

Version and dates: sample-plan-01. Publication, record-update and check dates are blank. No run has been recorded.

Decision to evaluate: Compare times within each room and keep incomplete bookings in a review list. The coordinator still decides whether to confirm them.

  • Same room, 09:00 to 10:00 and 09:30 to 10:30: expect an overlap flag. Actual result: not run.
  • Different rooms at the same time: expect no same-room conflict. Actual result: not run.
  • A booking with no end time: expect a review item. Actual result: not run.

Failures and limits: No observations recorded. The treatment of adjacent bookings and invalid times still needs a rule and a check. Coordinator acceptance is pending.

Handoff: The owner needs a way to inspect the flagged booking IDs, correct the fictional input and rerun the checklist. Recovery instructions have not been written or tested.

Evidence: The example has expected cases only. A completed record would link the actual run and review evidence, with publication permission checked separately.

Download a blank copy ↓