All products

Check data analysis quality with AI

Run a local check on an output file and pair the result with a human review of definitions, reasoning and presentation before a report leaves the team.

Check data analysis quality with AI when a plausible-looking chart might conceal a duplicate row or a missing value. The analysis-qa-checklist skill includes a local Python checker for the output dataset and a broader sign-off checklist. Some checks are mechanical; deciding what a flagged result means requires knowledge of the analysis.

An order date in the future could indicate a faulty import. In a booking dataset, a future date may be exactly what the field represents. The review can record that distinction, along with the owner who checked it, instead of silently dropping the row or waving the warning through.

What it does

  • Confirm the dataset has rows before interpreting summary figures.
  • Flag duplicated rows and significant missingness in columns.
  • Inspect numeric anomalies and date fields that deserve a second look.
  • Write a local JSON record of automated checks when needed.
  • Review metric logic and whether the written conclusion matches the evidence.

Install and use

  1. Install the skill in a compatible agent with Python 3 and pandas. Use a local CSV or a Parquet file with a supported pandas reader installed.
  2. Ask, “Check this analysis output, then separate file-level warnings from questions about the business definition.”
  3. Review warnings in context, record the treatment of each material issue and have the responsible reviewer confirm readiness.

FAQ

Source

Made by Nimrod Fisher. Original skill: analysis-qa-checklist, published on GitHub under the MIT licence.

View on GitHub