Report output validator
Skill nicholashidalgo/claude-skillforge/data-engineering/report-output-validator
npx -y skills add nicholashidalgo/claude-skillforge --skill report-output-validatorAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Validate report outputs with Python using row-count checks, subtotal tests, duplicate-grain checks, and comparison summaries.
SKILL.md
1.8 KB, as published. Nobody here has run it
Runtime Configuration
version: "1.0.0"
gotcha_pack: "sql-data-gotcha-pack"
gotcha_pack_version: "1.0.0"
gotcha_enforcement: "block_on_high"
Purpose
Validate report output before release or after logic changes.
Checks
- row counts
- duplicate grain checks
- subtotal to total tie-out
- period coverage
- category coverage
- comparison to prior or source extract
- reasonableness checks
Output format
- Validation scope
- Pass or fail checks
- Exceptions
- Python script
- Release risk summary
Gotcha Enforcement
Every validation script produced must satisfy these rules. HIGH violations block output. MEDIUM violations appear in Validation results as warnings with explanation.
| ID | Sev | Check |
|---|---|---|
| G002 | HIGH | If the report involves a join, state the expected cardinality and test for fan-out |
| G003 | HIGH | Every aggregation in validation code documents its NA/null behavior explicitly |
| G007 | HIGH | Validation logic must not replicate the report's own SQL or transformation logic |
| G009 | MEDIUM | Null rate for every critical measure column must be checked and surfaced |
| G012 | HIGH | Confirm grain, period, and filter alignment before comparing source to report |
| G014 | HIGH | At least one measure value must be reconciled; row count alone is not sufficient |
| G015 | MEDIUM | If net difference is zero, run a segment-level check before calling the report clean |