Feature audit found 11 things v2 dropped from v1. Restored the 7 that affect
output quality: quality checklist for prompts, anti-pattern examples, self-check
instruction, "don't re-search" context memory, prompt format anti-pattern,
multi-prompt guidance, and emoji footer with engagement counts.
Also adds visible query parsing display (🔍 **{TOPIC}** · {QUERY_TYPE}) and
the full v1-vs-v2 comparison test plan with all 17 test queries from README.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Matt Van Horn