Brief
An independent verifier clones your repo cold, builds the image, runs the containerized smoke suite, and confirms green — witnessed or scripted with logs.
Dataset
Your flagship repo at a pinned commit + the m0 reproducibility checklist, used here as the grading sheet (each of the 8 items pass/fail).
Starter guardrails (plan 88)
Provide REPRO.md (exact commands, expected outputs, timings); pin the commit
hash in the submission; no verifier-side prerequisites beyond Docker.
100-pt rubric
Checklist items 1-8 at 10 pts each (80); REPRO.md clarity 10; log evidence 10.
Deterministic checks -> LLM review
Checks: verifier log shows build success + pytest -q green; commit hash
matches; all 8 items marked. LLM review grades REPRO.md and failure candor.