Do I need an OpenAI Codex API key to use this?

While the skill can run in 'claude-only' mode, the advanced collaboration features require the Codex CLI (codex exec) to be available on your system.

Can this skill modify my source code during an investigation?

The skill may suggest instrumentation or tests, but it includes safety guards that require user confirmation before any file modifications or potentially destructive operations are executed.

What is the Strong Inference method used in this skill?

Strong Inference is a scientific approach that involves generating multiple competing hypotheses and designing experiments specifically to eliminate them until only the most likely explanation remains.

Where does the skill store its investigation findings?

All investigation states, evidence, and experiment logs are persisted to a structured markdown file located in the tmp/strong-inference/ directory for easy review.

Strong Inference Debugger

Name: Strong Inference Debugger
Author: masuP9

bymasuP9

•

セキュリティとテスト

Investigates complex software bugs and identifies root causes using a systematic, hypothesis-driven scientific methodology.

This skill implements the 'Strong Inference' scientific method to accelerate debugging by generating competing hypotheses and designing decisive experiments to eliminate them. It transforms vague troubleshooting into a structured process of elimination, allowing Claude to collaborate with OpenAI Codex for advanced hypothesis generation while maintaining a detailed investigation log. By focusing on proof and evidence rather than guesswork, it helps developers resolve intermittent errors, race conditions, and performance bottlenecks with high confidence.

主な機能

012 GitHub stars

02Experimental design focused on hypothesis elimination through 'killer experiments'

03Systematic hypothesis-driven debugging workflow

04Generation of mutually exclusive and testable competing hypotheses

05Collaborative mode leveraging OpenAI Codex CLI for deep analysis and review

06Persistence of investigation state in a structured markdown hypothesis tree

ユースケース

01Investigating intermittent or non-deterministic production errors

02Troubleshooting race conditions and state-related bugs in distributed systems

03Performing root cause analysis for complex system performance bottlenecks

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add masup9/codex-collab strong-inference

For use in Claude.ai and ChatGPT

Download Skill