# Worked synthetic annotation portfolio

Task: Apply LearnWithZavi guideline 1.0 to C05 C08 C10 N01 N08 R05.
Data origin: Existing LearnWithZavi synthetic stationery-support lab, 12 October 2026.
First pass: Deliberately invented teaching answers from the reviewed 15 October article. It contains three planned mistakes, not a learner performance record.
Results: First pass intent 1/3, whole entity records 1/2, reply ratings 1/1. Corrected 3/3, 2/2, 1/1 after key review. These mixed-task counts are not an industry accuracy score.
Correction example: N01 changes ORDER_ID "42" to exact text "0042". PRODUCT "blue notebook" and DATE_TEXT "tomorrow" stay unchanged. The source is not edited.
Evidence: [source records](../learner/sources.csv), [rules](../learner/rules.md), [fictional first pass](first-pass.csv), [corrected pass](corrected-pass.csv) and [decision log](decision-log.csv). These links are relative to this README. JSON alternatives accompany both passes.
Assistance and review: AI-assisted packaging and source checks; no completed human review or learner pilot.
Limits: Synthetic, six records, answer-key-exposed, no client, payment, independent assessment, full-lab completion or employment claim.