Identify the 20% of tests eating 80% of CI runtime — with concrete fixes.
When to use this prompt: Identify the 20% of tests eating 80% of CI runtime — with concrete fixes. Drop it into Claude Code (or paste at the top of a claude.ai conversation) when you hit the exact moment described — the template is structured so Claude returns a focused, scannable answer, not a wall of text.
What this prompt does
Frames the task with a clear single goal so Claude doesn't drift into adjacent work.
Constrains the output shape — bullets, ranks, or diff-only — so you can scan in under 30 seconds.
Asks for the assumptions Claude made, so you can correct course before the answer commits to a wrong frame.
Pairs with a CLAUDE.md auto-load so you don't have to paste it every time — save once, use every session.
Prompt structure (preview)
The actual prompt body (with the exact phrasing that makes Claude return cleanly) is in the Power Prompts PDF. The shape looks like this:
# Slow Test Speedup — Power Prompt #13ROLE:[role framing — full text in PDF]INPUT:[what you paste in]CONSTRAINTS:
- [constraint 1 — keeps output focused]
- [constraint 2 — output shape]
- [constraint 3 — what Claude must NOT do]OUTPUT FORMAT:[bullets / table / diff — exact spec in PDF]TIP:[battle-tested usage note — what to do when output is weak]
How to use it (3 steps)
Open the Power Prompts PDF, copy the prompt body for Slow Test Speedup.
Fill in the [BRACKETS] with your code, error, or context.
Paste into Claude Code or claude.ai — or save in CLAUDE.md so Claude auto-loads it every session.
Get the actual prompt body
30 copy-paste templates including Slow Test Speedup — 8-page PDF, paste into CLAUDE.md at your project root.
Is the full Slow Test Speedup prompt body free or paid?
It's part of the paid Power Prompts PDF (50p). Five other prompts are free on the free-prompts page — same format, so you can verify quality first.
Can I get a refund if Slow Test Speedup doesn't work for my workflow?
Yes — 7-day refund, no questions asked. Reply to the Stripe receipt email.
Will this work in claude.ai too, or only Claude Code?
Both. Every prompt is plain text — works in claude.ai, the Anthropic API, Claude Code, or any other Claude interface. The CLAUDE.md auto-load is Claude-Code-specific but optional.
How is this different from prompts I'd write myself?
The bracket-template structure (ROLE / INPUT / CONSTRAINTS / OUTPUT FORMAT / TIP) is the result of 50+ iterations across real codebases. Each prompt has a TIP block with what to do when Claude's first answer is weak — that's the part that takes hours to discover yourself.
💥 50p: 30 Claude Code prompts · £5£0.50Get the PDF →