SUBMISSION NOTES — Ibrutinib evidence audit v0.2.1

Upload 01_input_materials.zip as participant inputs and 02_reference_and_evaluation.zip
as organizer-only reference/evaluation dependencies. Only 01 is agent-visible.
Difficulty Self-Test is optional: leave its fields blank. No model performance
results, solver transcripts or self-test deliverables are included.

Changes from 0.2.0:
- K08 asks whether the included C records establish a measured whole-route yield
  including preparation of C_IV. Only not_established receives verdict credit.
- Report and claim explanations are presentation-only. Their automated checks
  assess submission format; scientific scoring applies to structured fields.
- The task contributor has confirmed a completed cross-check. The reference is
  recorded as contributor_cross_checked, with the exact statement and reference
  hashes in private/reference/review_record.json. The chosen reviewer label is
  'task contributor'. External independent expert approval is not claimed.

Molecular assignments, operations, route assessments, yield values, the remaining
claim verdicts, scoring weights and citation requirements are unchanged.

After extraction, enter ibrutinib_evidence_audit/ and use Python 3.12:
  python3.12 -m venv .venv
  .venv/bin/python -m pip install -r requirements.txt
  .venv/bin/python verify.py
  .venv/bin/python main.py grade "/absolute/path/to/agent/output" > ../score.json
Keep agent outputs and score.json outside the verified package root.

No .DS_Store, AppleDouble or __MACOSX metadata files are included. Manifests were
regenerated for this revision. Technical checks are not experimental validation;
no complete ALE Linux VM run or official qualification is claimed.
