Prompt·Accuracy and evidence·PG02
Find data that is internally consistent but wrong
Every check passes, but totals or outputs still look implausible.
Prompt
Look for data that passes the database's constraints and ordinary validation but is still analytically wrong. Test cross-table relationships, accounting identities, historical continuity, plausible ranges, source coverage, and independent calculations. Focus on contradictions that existing checks would miss, and distinguish genuine errors from legitimate differences in definitions or reporting periods.
- What it returns
- Cross-table reconciliations, identities and expected-population comparisons that fail, with source samples and how much data is affected.
- Skip it when
- There is no derived or analytical output, only records stored as entered.
- Avoid
- Calling different grains, as-of dates or reporting bases errors without checking their definitions.
- Changes
- Changes nothing
- Sessions
- Look first, then fix in a new session
- Tools
- Needs: the repository and read-only database access. Helps: pgTAP on a local or branch database, to turn confirmed rules into tests.
- Works on
- Any PostgreSQL host: Supabase, Neon, RDS, Cloud SQL or self-hosted.
Related prompts
- Missing vs zero vs stalePromptFinds where 'we don't know' turns into 0 and then into a wrong total.
- Rules the data never enforcesPromptFinds the rules everyone assumes (one owner per task, no negative balance) that nothing actually checks.
- Can we independently prove these numbers?PromptImportant figures may be wrong even though their lineage is documented.
- Find where records silently change identityPromptSources use aliases, changing identifiers, share classes or inconsistent names for the same thing.
More in Accuracy and evidence.