How a finding is traced to policy and evidence.
A 4-page report from the free sample audit. It follows one finding from the policy clause it breaks, through the test case that exercised it, to the recorded tool call and backend record, and shows how to reproduce every number.
- Pages 1–2. One finding traced end to end: policy clause, test case, tool call, backend record, cause and retest, plus a traceability matrix.
- Page 3. How the audit works: reviewed rules, test cases, decisions fixed before the agent runs, and the checks behind each finding.
- Page 4. How to reproduce the report in your browser, with the reproduction pack, or by hand.
Run the same sample
The report comes from the sample audit on this site. Run the sample audit, then open case refund-06a under Regressions to see the rule, the expected decision and both versions’ results.
Recompute every number
The reproduction pack contains the policy, all 50 test cases, both agent versions’ recorded results and a short checker. Unzip it and run python3 verify.py (Python 3.8 or later, no installs). It recomputes the decisions, pass/fail results and version comparison in the report.
Illustrative sample: a fictional refund policy, simulated agents and simulated backend records. No customer system or data was used. The checker is an independent re-implementation of the scoring for this sample, not the Policy Audit engine. Results cover the reviewed test cases only.