Brief
Log a 30-50 mock ladder across mentor, peer, and AI mocks with per-mock reflections and a weakest-skill trendline.
Dataset
Ladder tracker (94 sections 7-8): mock type, round, score/grade, reflection, retry outcome.
Starter guardrails (plan 88)
Use the tracker template; log within 24h of each mock; reflections name one fix + one drill; no backfilled batches.
100-pt rubric
Coverage (all round types x3 sources) 35; reflection quality 30; trendline + retry evidence 20; completeness (30-50 logged) 15.
Deterministic checks -> LLM review
Checks: >= 30 rows logged; all three sources present; reflections non-empty. LLM review grades trend honesty and readiness judgment.