diagnose

Diagnose bugs and performance regressions through reproducible feedback loops and falsifiable hypotheses.

1|Updated May 7, 2026
One-click install
npx skills add https://github.com/Soham407/studio-kickstart --skill diagnose-soham407
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: diagnose
Source: https://github.com/Soham407/studio-kickstart/tree/main/architecture/matt-pocock-diagnose
Command: npx skills add https://github.com/Soham407/studio-kickstart --skill diagnose-soham407

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

This Skill helps you systematically diagnose hard bugs and performance regressions when staring at code is not enough.

Core Features & Use Cases

  • Feedback loop first: Build an agent-runnable pass/fail signal (failing test, curl/HTTP script, CLI fixture diff, headless browser assertions, trace replay, or a minimal harness) before changing anything else.
  • Structured investigation phases: Reproduce, generate 3–5 falsifiable hypotheses, instrument with targeted probes, and then fix with a regression test where a correct seam exists.
  • Perf-aware measurement: Establish a baseline and bisect/measure first instead of relying on logs for performance regressions.
  • Cleanup and post-mortem: Remove debug instrumentation and document what would have prevented the issue, including architecture handoff when testing seams are missing.

Quick Start

Use it when you need to diagnose this by running a structured loop: reproduce the failure with a reliable feedback signal, rank falsifiable hypotheses, instrument a single variable at a time, fix the root cause, and lock it in with a regression test.

Frequently Asked Questions about diagnose

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

FAQPage Schema
How do I debug hard bugs when staring at the code isn't enough?

Debugging hard bugs requires a disciplined diagnosis loop. You reproduce the failure with a reliable feedback signal, generate falsifiable hypotheses, instrument a single variable at a time, fix the root cause, and lock it in with a regression test.

What's the best way to diagnose performance regressions without relying on logs?

Diagnosing performance regressions requires establishing a baseline and bisecting or measuring first. Instead of relying on logs, you instrument targeted probes to capture specific metrics, identify the bottleneck, and verify the fix.

How do I set up a reliable feedback signal for reproducing failing checks?

A reliable feedback signal for failing checks can be an agent-runnable pass/fail script. You build a failing test, CLI fixture diff, curl/HTTP script, or headless browser assertions before changing any code to ensure consistent reproduction.

Can I use Playwright for headless browser assertions during a debugging loop?

Yes, Playwright can be used to build headless browser assertions. This provides an agent-runnable pass/fail signal to reliably reproduce broken behavior or failing checks during the structured investigation phases.

What are the limitations of hypothesis testing when diagnosing broken behavior?

Hypothesis testing for broken behavior requires a reproducible failure mode to be effective. If you cannot establish a reliable feedback signal or reproduce the issue consistently, the structured diagnosis loop cannot proceed to the instrumentation and fixing phases.

Do I need to write a regression test seam to fix performance regressions?

Yes, fixing performance regressions requires a regression test seam where a correct seam exists. After instrumenting and measuring to find the root cause, you lock the fix in with a test and clean up any debug instrumentation.