Tanakh atlas receiving trial — complete research record

This download contains the sender's reading and rubric. For a fresh initial trial, use inputs/receiver-kit.zip and withhold the rest.

- inputs/: exact original receiver ZIP.
- returns/: exact returned ZIP, preserving original inputs, Python code, reports, interpretation and first-run history; the pasted response is also preserved separately.
- sender/: frozen composition, packet, exact expression, English reading and pre-trial rubric.
- publication-audit/: publisher's local reproduction, source and semantic assessments, and logs. These are not another external AI trial.
- sender-verification/ and published-snapshot/: what the sender checked before this trial, not perpetual current website state.
- results.txt / results.json: concise human-readable and structured publication records.
- FILES.json: byte counts and hashes of every other member.

Portable reproduction route: extract returns/receiver-return.zip, enter usl_tanakh_atlas_recovery, and follow METHOD.md. With local Python and Node:
python code/verify_inputs.py
python code/run_reference.py
python code/audit_tanakh.py
python code/build_reading.py
python code/inspect_passages.py
python code/test_audit.py

These commands regenerate data, not the authored English interpretation. The reference guard is an API guard, not an OS sandbox. Source HTML is data, not code to execute. Preserve the original archive.

Publication-tools retain their original release96/release97 workspace assumptions. They are an audit/build record, not portable installers. The returned kit above is the self-contained reproduction path.

One contextual receiving session is recorded. Its model label and past tool use are self-reported. Exact bytes, source correspondence, interpretation and agreement remain separate findings. The 79 new applications of older checked definitions remain proposed. No additional dictionary entries or revised atlas are silently introduced by this publication.
