💡 How to use these prompts:
Replace everything in [BRACKETS] with your specific details before sending.
Click Copy to copy any prompt to your clipboard instantly.
I want to accomplish this: [TASK]
Before doing anything, produce a plan that includes:
- The specific files you need to read, and why each one
- What you can do without reading
- Where you will need my input
- Roughly how many steps this will take
Wait for my approval. If the plan needs more than eight steps, propose a way to split it.
I need you to work on [TASK].
Before reading files, tell me which ones you think you need and why. I will confirm or correct the list.
Do not read the whole directory. Do not read files to build general context. Read only what the task requires, and say if you later need something you did not list.
We are partway through a long task. Before continuing, write a compact handover:
- What has been done
- Decisions made and why
- What remains
- Anything about this codebase worth carrying forward
Then I will start a fresh session with that summary instead of the full history.
Here are several small changes I need: [LIST]
Group them into the smallest number of passes, where each pass touches a coherent set of files and can be reviewed as a unit.
Tell me which groupings are safe to combine and which must stay separate because they could mask each other in review.
Here is what I use my coding agent for: [LIST]
For each item, tell me whether it needs a frontier model or whether something cheaper would do, and what specifically would degrade if I downgraded it.
Be concrete about the failure, not just that quality would drop.
Here is a description of how I use my coding agent in a typical week: [DESCRIBE]
Tell me:
- Which activities likely consume the most
- Which produce the least value per unit consumed
- The single change that would free up the most headroom
- What I should measure to confirm it worked
Give me one thing to change, not a list.