Previous version only looked at top-level metadata_element_list, which
contains categories — actual fields nest under each category. Now
recursively descends through all nested metadata_element_list arrays
and counts every element_id at any depth, then searches the full set
for CX/score/quality hints. Reports max nesting depth and the depth at
which each CX-flavored ID was found.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
psycopg2 performs %-substitution when params are passed to execute(),
so 'M%' in the LIKE clause was being interpreted as a positional
placeholder, raising IndexError when there's only one real %s (LIMIT).
Escape as 'M%%' so it's preserved as a literal percent.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Read-only script that samples B1 global masters from master_assets and
reports: top-level keys in full_metadata, presence of
metadata.metadata_element_list, and any element_ids matching
creativex/cx/score/quality (case-insensitive). Helps diagnose why the CX
backfill found 0 matches — distinguishes "client masters have no CX
score yet" from "CX field uses a different element_id than A1".
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
