diagnose

Diagnose bugs and performance regressions through a disciplined reproduce-and-measure loop.

2|Updated Jul 24, 2026
One-click install
npx skills add https://github.com/konstantinos-malavazos/claude-code-playbook --skill diagnose-konstantinos-malavazos
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: diagnose
Source: https://github.com/konstantinos-malavazos/claude-code-playbook/tree/main/templates/skills/diagnose
Command: npx skills add https://github.com/konstantinos-malavazos/claude-code-playbook --skill diagnose-konstantinos-malavazos

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When something breaks or slows down, the temptation is to guess at a fix and patch the symptom. This Skill replaces guess-and-patch debugging with a disciplined loop that proves the root cause from data before any code changes, so fixes actually stick and regressions get covered by tests. ## Core Features & Use Cases - Structured diagnosis loop: Reproduce the failure, minimise the repro, state a falsifiable hypothesis, instrument to confirm or refute it, then fix the cause rather than the symptom. - Performance regression workflow: Measure first to find the real hot path, compare against a known-good baseline commit or tag, fix the dominant cost, and re-measure to confirm the win. - Regression protection: Add a test that would have caught the bug, verifying the repro is gone and nothing else broke. - Use Case: A user reports an endpoint intermittently throwing 500s. Instead of patching the obvious line, the Skill guides a minimal repro, targeted logging, and a data-backed fix with a regression test. ## Quick Start Ask the assistant to diagnose this bug using the diagnose skill, starting from a minimal reproduction of the failure.

Frequently Asked Questions about diagnose

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

FAQPage Schema
How do I debug a bug I cannot reproduce?

Without a reliable reproduction you cannot claim a fix, so the first step is gathering more information until a minimal repro exists. The diagnose loop requires reproduction before hypothesising, and says so explicitly rather than guessing.

How to find the cause of a performance regression?

Measure first to locate the actual hot path with data rather than intuition. Compare against a known-good baseline such as a prior commit or tag, fix the dominant cost, then re-measure to confirm the improvement is real.

What is a falsifiable hypothesis in debugging?

It is a specific statement about the cause that instrumentation can confirm or refute. You add logging, assertions, or measurements that let the data decide, instead of pattern-matching to a usual suspect.

When should I not use a structured diagnosis loop?

The loop is designed for hard bugs and regressions where the cause is unclear. For trivial, obviously understood typos or misconfigurations, a full reproduce-instrument cycle adds unnecessary overhead.

Why fix the cause instead of the symptom?

Patching symptoms leaves the underlying defect in place to resurface elsewhere. The loop requires proving the cause from returned data or measurements, then making the smallest change that addresses it, plus a regression test.