Matt PocockStops the agent from guessing at a fix. It reproduces, narrows, forms one hypothesis at a time and tests it, which is the difference between a bug being fixed and a symptom being buried.
In the author’s words
Diagnosis loop for hard bugs and performance regressions. Use when the user says "diagnose"/"debug this", or reports something broken/throwing/failing/slow.
The description field from SKILL.md, quoted as written.
Read it before you install it
SKILL.md · 133 lines
Install on its own
Or add it to your stack above and install everything at once.
npx skills add https://github.com/mattpocock/skills --skill diagnosing-bugsTrust audit
View the source ↗- Checked automatically and confirmed.
- Resolved (HTTP 200), passed
- Checked automatically and confirmed.
- Resolves · 263,533 stars, passed
- Checked automatically and confirmed.
- Transcribed from the author's own page, passed
- Checked automatically and confirmed.
- Last updated 28 days ago, passed
- Checked automatically and confirmed.
- Credited to Matt Pocock, passed
Goes well with
systematic-debugging
Forces a method before any fix is proposed.
Fires on any bug, failing test or surprise, and it fires before the agent is allowed to suggest a change. The order is the whole point: understand the failure, then propose. It is the counterweight to an agent that is eager to be helpful.
npx skills add https://github.com/obra/superpowers --skill systematic-debugging
DebuggingFull page & trust audit →webapp-testing
Drives a local web app with Playwright.
Lets the agent open your dev server, click through it, screenshot what it sees and read the browser console. The gap it closes is an agent that writes frontend code it has never actually run.
npx skills add https://github.com/anthropics/skills --skill webapp-testing
Jesse Vincent
Anthropic