debug-hypothesis

Structure bug investigations with an Observe, Hypothesize, Experiment, and Conclude loop.

231|24|Updated Apr 11, 2026
One-click install
npx skills add https://github.com/LichAmnesia/lich-skills --skill debug-hypothesis-lichamnesia
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: debug-hypothesis
Source: https://github.com/LichAmnesia/lich-skills/tree/main/skills/debug-hypothesis
Command: npx skills add https://github.com/LichAmnesia/lich-skills --skill debug-hypothesis-lichamnesia

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Use when debugging any non-trivial bug — wrong output, crash, flaky test, performance regression, or "it works locally but not in CI." Forces a scientific-method loop (Observe → Hypothesize → Experiment → Conclude) so the agent stops guessing and starts reasoning. Prevents the #1 AI debugging failure mode — bulldozing through a wrong idea instead of falsifying it.

Core Features & Use Cases

  • Enforces a disciplined debugging workflow that records observations, hypotheses, experiments, and conclusions.
  • Prevents irrational fixes by requiring minimal, testable changes and evidence-based reasoning.
  • Applicable to CI failures, flaky tests, performance regressions, crashes, and unexpected behaviors across environments.

Quick Start

Begin with OBSERVE to collect facts, document them in DEBUG.md, and proceed through Hypothesize, Experiment, and Conclude.

Frequently Asked Questions about debug-hypothesis

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
What is hypothesis-driven debugging and when should I use it?

Hypothesis-driven debugging enforces an Observe, Hypothesize, Experiment, and Conclude loop for complex bugs like flaky tests or CI failures. It prevents guessing by requiring reproducible evidence gathering and minimal-experiment changes to falsify root causes.

How do I start a structured debugging workflow for a flaky test?

To start structured debugging for a flaky test, begin with the OBSERVE phase to collect all reproducible facts and document them in DEBUG.md. Then proceed through Hypothesize, Experiment, and Conclude to systematically track and validate the root cause.

Can I use this hypothesis debugging loop for CI environment failures?

Yes, this hypothesis debugging loop targets complex software bugs in both local and CI environments. It provides clear guardrails and documentation to investigate unexpected behaviors and environment-specific regressions through reproducible evidence gathering.

What is the best way to prevent AI agents from bulldozing wrong bug fixes?

The best way to prevent bulldozing wrong fixes is enforcing a scientific-method loop that requires minimal, testable changes. By tracking hypotheses and demanding evidence-based reasoning before concluding, the agent must falsify bad ideas instead of blindly applying them.

Why does my debugging process keep failing to find the root cause of performance regressions?

Debugging performance regressions often fails without structured hypothesis tracking, leading to irrational fixes. Applying a disciplined workflow that records observations and experiments ensures reproducible evidence gathering to accurately conclude the actual root cause.