Overview
Use this skill when correctness must be proven with control evidence, not only tests. It helps agents design reconciliations, control totals, exception workflows, and audit-friendly validation for high-trust datasets.
When to Use
- finance, billing, or revenue pipelines
- audit-sensitive operational datasets
- month-end or close-process data products
- source-to-target control validations
- exception-based review flows
Do not rely on a few spot queries when the business requires reconciled numbers.
Workflow
-
Define the control objective. Clarify:
- what must reconcile
- acceptable variance
- reconciliation frequency
- owner of exceptions
-
Choose the reconciliation pattern. Common patterns:
- row-count reconciliation
- control totals
- aggregate balance checks
- record-level exception matching
-
Make timing and cutoff rules explicit. Reconciliation often depends on accounting windows or source close timing.
-
Capture and route exceptions.
-
Preserve evidence. Control systems need reviewable records, not only ephemeral job output.
Common Rationalizations
| Rationalization | Reality |
|---|---|
| "The transformation logic is simple enough to trust." | Critical numbers still need independent validation evidence. |
| "A small variance is probably fine." | Acceptable variance must be defined, not guessed after a miss. |
| "The warehouse total matches once, so we are done." | Control reliability requires repeatable evidence over time. |
Red Flags
- no explicit acceptable variance exists
- cutoff timing is undocumented
- exceptions are noticed manually and inconsistently
- control evidence cannot be reproduced later
Verification
- Control objectives and acceptable variance are defined
- Reconciliation logic is explicit and reviewable
- Exceptions have an owner and workflow
- Control evidence is retained for audit or review