Skill · Prompt

Diagnose From the Symptom, Not the Guess

A prompt that forces a real diagnosis before a single line of fix.

TypePrompt

Something's broken. You describe it; the agent immediately proposes a fix. You try it — still broken. It proposes a different fix. Still broken. Forty minutes in, you have four patches, three new quirks, and the original bug — because nobody ever stopped to look at what was actually failing.

Every one of those patches was aimed at a guess. The agent pattern-matched your description to a plausible cause and changed code to fix that imagined cause — never the real one. A diagnosis takes effort and a guess is free, so it reaches for the guess. This prompt makes it earn the fix:

Without the prompt

Hears the description, patches the likely cause
Still broken → patches again → again
Four patches, new quirks, original bug intact

With the prompt

Reproduces the symptom, reads the real output
Names the root cause and the evidence for it
One fix, aimed at the actual cause

Paste it when something is failing and you want the cause, not a lottery. Its sharpest moment is the second "still broken" — that sentence is information: it means the diagnosis was wrong, not the patch under-applied. The prompt turns it into a re-read of the new output instead of another roll of the dice.

Paste into the chat
Before proposing any fix, diagnose first:
1. Reproduce the symptom and paste the actual error or wrong output — the real text, not a paraphrase.
2. Trace it to a root cause. State the cause and the specific evidence that points to it.
3. Only then propose the smallest change that addresses that cause.
If I say it's still broken, do not try a different patch — re-read the new output and re-diagnose. A second failed fix means the first diagnosis was wrong.
Prefer your agent install it?
Did this help?
Other skills
c
The author

Each story here wraps a lesson paid for in full.

craftagentsomeone building and learning at once

What are you building with agents? Want to trade notes, push back, or build something together — drop a line.

58pieces12clustersVI·ENbilingual

Get new pieces by email

Field notes on working with AI agents — occasional, no spam.