ANONYMOUS SUPPLEMENTARY CODE AND DATA

This archive supports the controlled validation manuscript submitted to AI and Ethics. It contains the executable reference implementation, the pre-run source configuration, the frozen results-directory configuration, episode- and step-level outputs, statistical summaries, threshold sensitivity, computation accounting, run manifest, figures, and SHA-256 checksums.

The archive contains no author names, affiliations, contact details, or local absolute paths. The hidden evaluator is used for post-action evaluation and is not exposed as an instance-level policy label.
