validate-data
QA an analysis before sharing -- methodology, accuracy, and bias checks. Use when reviewing an analysis before a stakeholder presentation, spot-checking calculations and aggregation logic, verifying a SQL query's results look right, or assessing whether conclusions are actually supported by the data.
Security Assessment
About validate-data
The validate-data skill provides comprehensive quality assurance for data analyses before they are shared with stakeholders. It systematically reviews methodology, calculations, and conclusions to catch errors, identify biases, and assess whether findings are truly supported by the data. This skill is designed to prevent the costly mistakes that occur when flawed analyses reach decision-makers, ensuring that presentations, reports, and recommendations are built on solid analytical foundations.
The skill performs multi-layered validation including methodology review (question framing, data selection, metric definitions), calculation verification (spot-checking numbers, validating aggregations, confirming comparison logic), and pitfall detection (join explosions, survivorship bias, timezone mismatches, selection bias, and other common analytical errors). It also evaluates visualizations for misleading elements like inappropriate axes or distorted scales, and assesses whether narratives and conclusions logically follow from the evidence presented.
This skill is valuable for data analysts preparing stakeholder presentations, anyone reviewing SQL query results before sharing insights, teams conducting pre-delivery quality checks on reports, and professionals who need to verify that their analytical conclusions are defensible. It generates a confidence assessment with one of three ratings—ready to share, share with noted caveats, or needs revision—along with specific, actionable improvement suggestions to strengthen the analysis.
FAQ
What types of analyses can this skill validate?
The skill can review various analysis formats including documents and reports in conversations, files (markdown, notebooks, spreadsheets), SQL queries with their results, charts with underlying data, and descriptions of methodology and findings.
What does the confidence assessment rating mean?
The skill provides one of three ratings: 'Ready to share' means the analysis is methodologically sound with only minor suggestions; 'Share with noted caveats' means the analysis is largely correct but has specific limitations that must be communicated; 'Needs revision' means specific errors or methodological issues were found that should be addressed before sharing.
What kinds of errors does this skill check for?
The skill checks for methodology issues (wrong data selection, inappropriate time ranges, unclear metrics), calculation errors (incorrect aggregations, mismatched totals, wrong comparison periods), common pitfalls (join explosion, survivorship bias, timezone mismatches, selection bias), and presentation problems (misleading visualizations, unsupported conclusions, missing caveats).
Do I need to provide the underlying data for validation?
Not necessarily. The skill can work with the analysis as presented—whether that's a report with findings, SQL queries with results, or charts with data. However, having access to underlying data or queries allows for more thorough spot-checking and independent verification of calculations.
Can this skill fix errors it finds, or just identify them?
The skill focuses on identification and assessment, providing specific actionable suggestions for improvement. It will point out what needs to be corrected, suggest additional analyses that would strengthen conclusions, and recommend caveats that should be noted, but the actual corrections are left to the user.
Install validate-data
Quick Setup:
- Copy the skill folder to
.claude/skills/ - Claude will automatically detect and use the skill
Repository
anthropics/knowledge-work-plugins