What is the main philosophy of DS Fix?

It follows the 'Iron Law of DS Editing,' which mandates diagnosing the root cause before changing any code to prevent cosmetic patches and recurring bugs.

Is it integrated with other workflow documentation?

Yes, it automatically reads SPEC.md, PLAN.md, and LEARNINGS.md to ensure any fix aligns with the original project objectives and previous findings.

Can it handle SAS projects?

Yes, it includes specific detection and enforcement for SAS ETL failure modes like hash merge memory issues and SGE array misconfigurations.

What happens if my source data changes mid-project?

The skill triggers a Re-profiling Protocol to compare new data shapes, dtypes, and distributions against original baselines in LEARNINGS.md to identify shifts.

How does it handle incorrect analysis results?

It utilizes a 'Trace Backwards' protocol to find the first step where output diverges from expected values rather than guessing forward from the end result.

DS Fix: Data Science Troubleshooting

Name: DS Fix: Data Science Troubleshooting
Author: edwinhu

byedwinhu

•

Data Science & ML

Provides a structured diagnostic framework for fixing broken data science workflows, notebook errors, and incorrect analysis results.

DS Fix is a specialized Claude Code skill designed for mid-analysis course correction in complex data science projects. It enforces a strict 'diagnose before fixing' policy to move beyond guesswork, offering specific protocols for runtime errors, unexpected results, changing data schemas, and reviewer feedback. By integrating with existing project context like SPEC.md and LEARNINGS.md, it ensures that fixes address root causes rather than just symptoms, maintaining the integrity of the entire analytical pipeline through output-first verification.

Key Features

01Automated data re-profiling for handling schema, volume, or distribution changes

02Competing Hypothesis Investigation for resolving mysterious or contradictory bugs

03SAS-specific troubleshooting for enterprise ETL and memory-intensive workflows

045 GitHub stars

05Specialized protocols for runtime errors, wrong results, and reviewer feedback

06Strict 'Diagnose Before Fixing' enforcement to eliminate trial-and-error guesswork

Use Cases

01Correcting statistical results that diverge from business expectations or sanity checks

02Adapting an existing analysis pipeline to updated source data or new project requirements

03Debugging Python or SAS notebook cell failures and complex tracebacks

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add edwinhu/workflows ds-fix

For use in Claude.ai and ChatGPT

Download Skill