MICKAI TRAINING ACADEMY
Merefield administrative handover lab

Entirely fictional educational material. No real patient, staff, dispatch, vehicle or clinical information. No accreditation, operational permission or professional approval is granted.

1. Read charter.json, tasks.json and evidence.json. The cutoff and freshness interval are local learning rules.
2. Complete worksheet.txt and activities-and-answers.json on paper before opening the worked key.
3. Inspect checker.py and test_checker.py. The standard-library script reads local files and prints JSON. It has no network, model, message or source-writing function.
4. Optional Python 3 commands from this extracted folder:
   python checker.py flawed-handover.json
   python checker.py corrected-handover.json
   python -m unittest -v test_checker.py

Expected: flawed gives 21/30 matching fields, five defect records and five supplied rows. Corrected gives 30/30 and six supplied rows, but five tasks remain unresolved. Thirty tests include all 720 row-order permutations. These are local fixture checks, not model measurements.

checker.py compares a candidate with the separately authored expected.json. It does not calculate the correct operational status from raw evidence or validate a real reporter. Missing rows lose five comparisons each; duplicate/unknown task IDs, duplicate/unknown evidence IDs, malformed fields or a different cutoff reject structurally. Evidence set order is irrelevant. Keep the key unchanged while checking your answer. A key correction needs its own evidence and review.

Make one mutation in a copy and predict the result. Keep the original and your correction note. Do not introduce real data, external integrations or credentials. A manifest checks byte consistency against its recorded digests, not truth, identity or human approval.
