Full traces for the 25 max-effort runs behind the README results (99.86% mean RHAE, official scorecard 9c403765-db5b-40b1-beab-6fa3f40119b0), plus the superseded first attempts for tn36 and bp35 disclosed in the Validity section. Each run directory contains workspace/log.txt (every action, board, and level counter), workspace/playbook.md, workspace/scratch/, transcript.jsonl (every model invocation's full input and output), traces/ (per-request token usage), metrics.json, and containment.json/opening.png/prime.json.