What problem does it solve? After a calibration round clears its human gate, nothing in the workflow scheduled the curation of that round's verified findings, staged proposals, script templates, open questions, and provisional bounds into the site knowledge base, so rounds repeatedly lost their lessons and repeated their errors. ## Core Features & Use Cases - Sequenced post-round checklist: Runs seven ordered steps — knowledge curation, proposal promotion or discard, script review, open-questions emission, bound-debt recording, knowledge-base assertions, and state recording — dispatching each to the skill or tool that owns the procedure. - Non-empty knowledge assertions: Verifies all four knowledge stores (experiments, failed_approaches, discoveries, parameters) plus the run ledger are populated where the round produced findings, catching silent curation skips. - Fuller convergence checklist: On a CONVERGED round, adds six campaign-close items including the final reproducible configuration artifact, generalizability promotion, bound-debt discharge, unscored-data reporting, campaign ledger, and residual statement. - Use Case: After a calibration round's gate clears and resolve_next_action() returns close("housekeeping"), run this skill to curate the round's Phase-5-verified findings into use_cases/{Model}_{Case}/memory/gained_knowledge/ before the next Phase 0 begins. ## Quick Start Run the round housekeeping to close out this calibration round and curate its verified findings into the site knowledge base.
