What happens if a commit cannot be built or tested?

The skill includes a 'skip' strategy. If a specific commit is untestable due to build failures or environment issues, it can mark that commit as skipped and continue searching other parts of the history.

How does it help after finding the bad commit?

Once the 'first bad commit' is identified, the skill automatically hands off the context to systematic-debugging tools to analyze the diff and determine the actual root cause of the failure.

Does it work for UI and visual regressions?

Yes, by using the 'Manual' or 'Hybrid' strategies, Claude can provide specific verification steps for you to follow, or use subagents to inspect visual states at each commit.

Can I use this for automated testing?

Yes, the skill supports an 'Automated' strategy where you can provide a test script (bash, npm, etc.). Claude will automatically run this script at each step of the bisect to find the breaking change.

What is the git-bisect-debugging skill?

It is a specialized Claude Code skill that implements a binary search algorithm through your git history to identify the specific commit where a bug or regression was first introduced.

Git Bisect Debugging

Name: Git Bisect Debugging
Author: sjungling

bysjungling

•

開発者ツール

Systematically identifies the exact commit that introduced a bug or regression using a structured binary search workflow.

This skill automates and guides the process of finding regressions in your codebase by leveraging the power of git bisect combined with Claude's subagent architecture. It provides a rigorous four-phase approach—Setup, Strategy Selection, Execution, and Analysis—to ensure a clean state, verify good and bad commits, and execute tests across the commit history using automated scripts, manual verification, or a hybrid of both. By isolating test execution in subagents and handing off findings to systematic debugging tools, it transforms tedious 'when did this break?' investigations into a predictable, deterministic process.

主な機能

01Mandatory safety checks to verify clean working directories and valid commit ranges

02Isolated subagent execution for reliable commit verification

03Seamless handoff to systematic-debugging for post-bisect root cause analysis

048 GitHub stars

05Interactive progress tracking with standardized TodoWrite checklists

06Automated, manual, and hybrid bisect strategies for flexible debugging

ユースケース

01Pinpointing the specific commit that introduced a performance regression

02Debugging 'it used to work' scenarios in large, complex repository histories

03Identifying when a previously passing unit or integration test started failing

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add sjungling/sjungling-claude-plugins git-bisect-debugging

For use in Claude.ai and ChatGPT

Download Skill