Pattern Assimilation Step
Verifies each landed task filed exactly one learning record naming what it changed.
Contract
Does When a piece of work lands in the local system, this component checks the completion records for it — confirming that exactly one same-lane "what did we learn from this" decision was filed per landed item, that any claimed refinement names an owner-visible surface and the artifact it changed, that a "nothing to refine" entry carries its required typed fields, and that there are no duplicate or off-lane entries. It runs over fixture data and makes the completion bookkeeping inspectable, showing whether the recorded completions conform to the system's stated learning-from-landed-work rules, rather than leaving that on faith.
Scope limit It validates only the declared public completion contract over synthetic fixture data; it does not ingest private lessons, mutate live ledgers, promote global doctrine, include launch operations or public sharing, make external model access, claim non-public data equivalence, or certify public runtime behavior.
Run
PYTHONPATH=src python3 -m microcosm_core.validators.acceptance --only pattern_assimilation_step --input fixtures/first_wave/pattern_assimilation_step/input --out receipts/first_wave/pattern_assimilation_acceptance.json
Evidence
ClassContract validatorevidence 5/5Import validation
Source
Paper module
Paper module Pattern Assimilation