Anomaly Swipe
AI-powered anomaly detection with intuitive swipe-based auditor review, powered by a privacy-preserving static HTML viewer.
How It Works
Anomaly Swipe uses AI to analyze your data for quality issues, then presents the results in a swipe-based triage interface. Swipe right to investigate, left to dismiss. The generated HTML runs entirely in your browser—export your classified results for workpapers.
Tutorial: Detect Anomalies with Claude Code
Collect the data you want to analyze for anomalies—transaction logs, expense reports, journal entries, or any tabular data. Export it as CSV or Excel.
Download the Anomaly Swipe skill file and place it in your Claude Code custom skills directory. The skill will be available as a slash command.
Point the skill at your dataset. It will analyze the data for length anomalies, encoding errors, PII exposure, duplicates, pattern matches, and statistical outliers, then generate a standalone HTML triage interface with all anomalies embedded.
Open the generated HTML file. Each anomaly appears as a card. Swipe right (or click thumbs up) to mark for investigation. Swipe left (or click thumbs down) to dismiss. Use keyboard arrows for even faster triage.
When you've swiped through all anomalies, the results page shows a summary. Click "Export JSON" or "Export CSV" to download your classified anomalies for workpapers.