CawPlan Internal QA — Human Input Sample & Label

Bootstrap

cawplan skill check

Task

Fetch human-input rows via the CawPlan CLI, emit a fixture JSON aligned with production classify context rules, then open the local label editor so a human can set expected_categories, expected_topic, optionally expected_reason, and review_required per turn.

Workflow

1. Resolve scope (same disambiguation rules as humaninputs-test)