# Public projection: privacy and provenance

This public report is a sanitized projection of a local AIL-546 Eval study.

## Published

- the self-contained interactive report;
- the Markdown analysis;
- aggregate before/after machine results;
- the catalog of public professional-person examples;
- an aggregate readiness summary for the internal real-task catalog.

## Deliberately not published

- internal user and task identifiers;
- internal task titles and person names;
- blocked or unreadable task lists;
- authenticated dashboard URLs;
- raw task bundles, prompts, screenshots, Provider payloads, emails, or phone numbers;
- the local replay fixture, because some synthetic records preserve names derived from internal tasks;
- repository source and tests outside this report boundary.

The full local evidence remains reproducible inside the authorized repository. The public machine
result retains the baseline revision, scorer hashes, fixture hash, case IDs, metrics, and per-case
decision traces, but does not expose the private source rows used to derive the synthetic replay.

## Interpretation

The 11-case replay is diagnostic. It demonstrates deterministic scorer semantics and does not claim
live Provider recall, production accuracy, LLM routing quality, or current-role freshness.