Quality scenarios and executable evidence

This required 18–26-hour package strengthens guided Labs 1 and 3. Allocate 4–6 hours to lessons and predictions, 8–12 to evaluator implementation and tactic experiments, and 6–8 to assessment and repair. Reuse these outputs in the existing labs. The original eight scenarios, architecture views, three fitness functions, reversibility register and tradeoff review remain required. Estimates are author planning estimates; Semester 07 reconciliation follows its module audits.

Prerequisite bridge

Review ratios, inequality boundaries, sampling limitations and controlled comparisons from Semester 01. Revisit the Semester 06 distinction between a model check and evidence from an implemented service. Be able to distinguish a timeout, failed response, slow success and absent observation.

Sequence

1. Scenario contracts.
2. Evidence evaluator.
3. Tactic experiments.
4. Implement quality-evidence-contracts.zip from its README using Python 3 standard library only.
5. Complete QA1–QA6 assessment and calibration.

Run python3 -B check.py starter before opening the reference. The reference is calibration, not a submission. Fixtures test an offline evaluator; the original tactic experiment still requires actual implementations and measured observations. Preserve raw inputs, hand predictions, failed attempts, fixes and assistance disclosures.

Evidence limits

The harness checks 2,500 policy cases and their reversed input orders, time boundaries, sample minima and invalid observations. It cannot detect attempts omitted by an upstream collector. No real service load test or production SLO is claimed by the reference checks.

Semester integration

The reconciled workload now includes this package, original module work, required strand and assessment. Its allocation supersedes earlier provisional timing notes; do not count reused artifacts twice.