Module finalize 

Grading finalize.

For each (eval, condition) it grades transcript_check assertions directly, folds in the llm_judge responses written by the orchestrator (missing → FAIL), assembles the skill-invocation meta result, and writes a schema-valid grading.json with pass/fail summaries.

Structs

FinalizeSummary

What finalize graded, for the CLI summary.

Functions

finalize

Fold judge responses + transcript checks into a grading.json per (eval, condition). See the module docs for the per-assertion behavior.