bilingual-audit

Scan data/ fixtures for English-without-Malayalam parity gaps (titleMl, definitionMl, nameMl, summaryMl). Use when reviewing data changes, before PR, or when asked about bilingual coverage.

SKILL.md

GitHub

Bilingual audit

Reports every record in data/ where an English field is filled but its Malayalam (*Ml) counterpart is missing or empty. The product promise is bilingual parity — *Ml fields are typed as optional, so the type system won't catch gaps.

Fields to audit

File EN field ML field
data/kpis.ts title titleMl
data/kpis.ts meta.definition meta.definitionMl
data/departments.ts name nameMl
data/ministers.ts name nameMl

(If new fields with Ml siblings appear — e.g. summary/summaryMl — add them to the table.)

How to run

Quick grep pass:

# Find KPIs missing titleMl
grep -nE '^\s*title:' data/kpis.ts | while read line; do
  lineno=$(echo "$line" | cut -d: -f1)
  # check next ~3 lines for titleMl
  if ! sed -n "$((lineno+1)),$((lineno+3))p" data/kpis.ts | grep -q 'titleMl:'; then
    echo "data/kpis.ts:$lineno missing titleMl"
  fi
done

For a structured pass, read each fixture file end-to-end and inspect every record object — grep misses nested meta.definition cases. The fixture files are short (<100 records each); a full read is fine.

Output format

Group findings by file. For each gap, report:

data/kpis.ts: id=fiscal.revenue-deficit missing meta.definitionMl
data/departments.ts: id=dept.tourism missing nameMl

End with a one-line summary: N gaps across M records.

What to do with findings

This skill reports — it does not patch. Translations need a Malayalam speaker (or vetted glossary), not a guess. Flag the gaps and let the user decide whether to:

  1. File translation requests
  2. Mark the record dataStatus: "tbd" until the translation lands
  3. Remove the EN-only field if it was added prematurely

Never machine-translate Malayalam government terminology — wrong term = wrong meaning, and this dashboard is read by journalists.