Can I use this with other AI coding tools?

Yes, while designed for Claude Code, this protocol is compatible with any coding agent including Cursor, Codex, and OpenCode.

What is the 'Iron Law' in this protocol?

The Iron Law mandates that no code changes or fixes can be attempted until a full root cause investigation is completed and the failure is understood, preventing superficial 'symptom fixes'.

Does this skill replace existing testing workflows?

No, it complements them. It specifically triggers when tests fail or bugs are reported, acting as a bridge between identifying a failure and implementing a successful TDD fix.

How does the 'Three Strikes Rule' work?

If three consecutive fix attempts fail, the agent must stop, document the failures, and re-evaluate underlying assumptions or design patterns rather than continuing to try random changes.

Debugging Protocol

Name: Debugging Protocol
Author: jwilger

byjwilger

•

安全与测试

Implements a disciplined 4-phase debugging workflow that enforces root cause analysis and hypothesis testing before any code changes.

The Debugging Protocol skill transforms how AI agents handle errors by replacing guesswork with a systematic, scientific approach. It guides the agent through four distinct phases: gathering facts and reproducing errors, comparing failing code against working examples, testing specific hypotheses one change at a time, and finally implementing a verified fix. By enforcing the 'Iron Law' of investigation and the 'Three Strikes Rule' for escalation, this skill prevents common debugging pitfalls like symptom-patching and random 'trial-and-error' coding, ensuring that bugs are truly understood and permanently resolved.

主要功能

01The 'Iron Law' enforcement: Investigation is mandatory before any fix is attempted.

022 GitHub stars

03Single-change hypothesis testing: Requires undoing changes completely if a hypothesis is refuted.

04Three Strikes escalation: Prevents wasted effort by stopping after three failed fix attempts to re-evaluate assumptions.

05Root cause verification: Ensures fixes are backed by failing tests that now pass.

06Four-phase systematic debugging framework: Understand, Compare, Test, and Verify.

使用场景

01Tracking down regressions where previously working features have broken due to environment or dependency changes.

02Investigating intermittent test failures or unexpected runtime errors in complex systems.

03Systematically resolving user-reported bugs that require deep data-flow and stack trace analysis.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jwilger/agent-skills debugging-protocol

For use in Claude.ai and ChatGPT

Download Skill