diagnose

Diagnose hard bugs and performance regressions with a reproduce-minimise-hypothesise-instrument-fix-regression-test loop.

4|Updated Mar 2, 2026
One-click install
npx skills add https://github.com/shushuzn/Rairos --skill diagnose-shushuzn
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: diagnose
Source: https://github.com/shushuzn/Rairos/tree/main/skills/diagnose
Command: npx skills add https://github.com/shushuzn/Rairos --skill diagnose-shushuzn

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

Helps you systematically diagnose hard bugs and performance regressions when ad-hoc guessing wastes time and fixes don’t stick.

Core Features & Use Cases

  • Feedback-loop first: builds a fast, deterministic pass/fail signal so bisection, hypothesis-testing, and instrumentation can converge.
  • Reproduce with confirmation: verifies the failure matches the user-reported symptom and is repeatable enough to debug.
  • Rank falsifiable hypotheses: produces 3–5 testable guesses with explicit predictions before changing anything.
  • Targeted instrumentation: probes one variable at a time, tags debug logs, and focuses on boundaries that distinguish hypotheses.
  • Fix with regression coverage: turns the minimized repro into a failing test at the correct seam (or documents when no seam exists).

Quick Start

Ask an agent to diagnose the issue by constructing a fast repro feedback loop, reproducing the failure, ranking falsifiable hypotheses, instrumenting to test them, and finishing with a regression test or an explicit note that no correct test seam exists.

Frequently Asked Questions about diagnose

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

FAQPage Schema
How do I systematically debug hard bugs instead of ad-hoc guessing?▼

To debug hard bugs systematically, enforce a reproduce-minimise-hypothesise-instrument-fix-regression-test loop, building a fast pass/fail signal so bisection and hypothesis-testing converge quickly.

What's the best way to debug non-deterministic failures that are hard to reproduce?▼

Debug non-deterministic failures by increasing reproduction rate via stress or higher-trigger loops, then construct a reliable pass/fail signal to confirm the failure matches the user-reported symptom.

How do I rank hypotheses during debugging to avoid wasting time?▼

Rank falsifiable hypotheses during debugging by producing 3–5 testable guesses with explicit predictions before changing code, ensuring each guess targets boundaries that distinguish competing explanations.

How do I use targeted instrumentation to find a performance regression?▼

Use targeted instrumentation to probe one variable at a time for performance regression, tagging debug logs to focus on boundaries that distinguish hypotheses and validate the fix with a regression test.

When should I not use a disciplined debugging loop for issue diagnosis?▼

A disciplined debugging loop is not suitable when you cannot construct a reliable pass/fail signal or increase reproduction rate, as the loop requires repeatable failures to converge on a fix.

How do I create regression test coverage after fixing a hard bug?▼

Create regression test coverage by turning the minimized reproduction into a failing test at the correct seam, or explicitly documenting when no correct test seam exists for the fix.