Prompt·Accuracy and evidence·PG03
Find where records silently change identity
Sources use aliases, changing identifiers, share classes or inconsistent names for the same thing.
Prompt
Trace how real-world entities are identified, matched and connected across imports, source documents and database tables. Find cases where distinct entities are merged, identical entities are duplicated, or historical relationships become disconnected. Identify the matching rule responsible and the affected population before proposing corrections.
- What it returns
- Likely duplicates and false merges with their source IDs, the matching rule behind each, the records affected and a counterexample you can rerun.
- Skip it when
- Every source shares one stable identifier and nothing is matched by name.
- Avoid
- Merging entities just because their names look alike.
- Changes
- Changes nothing
- Sessions
- Look first, then fix in a new session
- Tools
- Repository and read-only database access.
- Works on
- Any PostgreSQL host: Supabase, Neon, RDS, Cloud SQL or self-hosted.
Related prompts
- Links between records worth addingPromptFinds links between records worth adding, and ones the evidence does not support.
- Imports that fail quietlyPromptFinds rows that go missing or land on the wrong record during import, with no error.
- Meaning drift across the stackPromptShows where one word (a task, a fund, a run) means slightly different things in different parts of the app.
- Find data that is internally consistent but wrongPromptEvery check passes, but totals or outputs still look implausible.
More in Accuracy and evidence.