[amar-python]

What

A sample Migration Verification Report (docs/samples/MIGRATION_VERIFICATION_REPORT_SAMPLE.md) produced by a real run, plus the script that produces it and the raw evidence behind it.

Refs #59. Not Closes: whether this meets "a Defence buyer could review" is the product owner's call, and the repo rule is that issues close on clean evidence.

File	Role
scripts/sample_verification_run.py	Drives the running API, cross-checks it against the database, renders the report
docs/samples/orgs_sample.csv	Synthetic T&E input: 10 rows, 4 deliberately defective
docs/samples/equipment_register.csv	Synthetic dynamic-mode input: 8 clean rows
docs/samples/verification-run-evidence.json	Raw API responses and database reads behind every figure
docs/samples/MIGRATION_VERIFICATION_REPORT_SAMPLE.md	The report

"Every figure traceable to a real run"

Nothing in the report is typed by hand. The script renders it from the evidence it records. The run was against PostgreSQL 16.15, the live API (real sign-in, real tenant schema) and a fresh database, at code version 01602c1. Result: 12 of 12 checks pass (6 of 10 T&E rows loaded, 4 rejected exactly as designed; 8 of 8 dynamic rows loaded; duplicates refused; audit chain intact).

The checks can fail

Re-running the script against the now-populated database exits 1 and writes "NOT RECONCILED" with 3 failed checks. I did that before committing, so a green report is not the only outcome the harness can produce.

What the report deliberately does not claim

Section 8 of the report says so in its own words: synthetic data only, development configuration (observed during the run: unauthenticated requests return 200, and the API connects as the audit table's owner), developer-run rather than independent, no ASDEFCON/DID claim (#159 is still open), and re-upload in T&E mode is unsafe (#215).

Found along the way

• T&E re-upload that fails every row still overwrites the earlier load's registry record #215: a T&E re-upload that fails every row still overwrites the registry record of the earlier successful load. RV-F3 (BR-34): all-failed T&E import runs registry cleanup and removes prior dataset #170's guard skips re-uploads. Reproduced on current main.

Reviewer notes

• Not run in CI. The script needs a live API and database, so it is a manual, repeatable procedure (steps in the script's docstring), not a test.
• Markdown lint: markdownlint-cli2 reports 0 errors on the report.
• Outputs contain run timestamps, so regenerating produces a different file; the counts and hashes are stable.

---

Generated by Claude Code