Bits of .NET

🤖 “It’s Broken” Isn’t a Bug Report — This Prompt Fixes That

A bug report that says “the checkout page doesn’t work sometimes” costs a developer real time just to reproduce, before they can even start fixing anything. Turning a vague report into concrete, testable repro steps is exactly the kind of structured, exhaustive work an AI prompt handles well.

🎯 Why Vague Reports Are So Expensive

"Sometimes" and "doesn't work" hide real information the reporter
DOES have but didn't think to include: what they clicked, what
device/browser, what data they were using, what they expected vs
saw. A developer has to go back and forth asking questions before
they can even attempt to reproduce it - often losing the reporter's
attention entirely by message three or four.

✅ The Prompt

• Paste the raw bug report exactly as the user/reporter wrote it — don’t clean it up first, the model needs to see what’s actually missing.

🤖 Copy This Prompt

Here is a raw, informal bug report. Turn it into a structured
report a developer can act on immediately:

1. SUMMARY: one sentence, specific, no vague words like
   "sometimes" or "doesn't work" - if the report doesn't specify
   something, say "NOT SPECIFIED" rather than guessing.
2. STEPS TO REPRODUCE: numbered, concrete steps, inferred as
   precisely as possible from what's given.
3. EXPECTED vs ACTUAL: what should have happened vs what did.
4. MISSING INFORMATION: a specific list of questions to ask the
   reporter to fill the gaps (browser/OS/device, exact input used,
   how often it happens, any error message text, screenshots).
5. LIKELY SEVERITY: your best guess (blocking / major / minor /
   cosmetic) with one sentence of reasoning.

RAW BUG REPORT:
[PASTE THE REPORT AS WRITTEN]

⚠️ This Doesn’t Replace Talking to the Reporter

• The “missing information” list IS the value here — it turns a vague back-and-forth into one targeted follow-up message instead of five.
• Don’t treat the inferred repro steps as confirmed until someone actually verifies them — they’re a strong starting hypothesis, not a guarantee.