Most AI advice online is either vague (“use AI to boost productivity!”) or a list of 47 magic prompts that never work outside a demo. This article is different. These are prompts I use daily as a freelancer and developer — they’re boring, specific, and they cut real hours off my week. No hype, just copy-paste and adapt.
The Customer-Reply Prompt That Ends Back-and-Forth
You’ve written a quote, and the client asks the same three questions: “Can you do it faster? Is that fixed price? What’s included?” Instead of typing a fresh email, use this:
You are a freelance project manager. Rewrite my draft reply to a client.
Keep the tone professional but warm.
Answer these implied questions explicitly: (1) timeline flexibility, (2) fixed vs hourly pricing, (3) exact deliverables.
Use plain English, no marketing fluff.
Draft:
[Paste your draft]
Why this saves hours: it forces you to write the draft once (even badly), and the AI structures it into a clear, client-ready response. You stop rewriting the same email five times. I use this for every quote, revision request, and scope change.
The “Tighten This Code” Prompt for Code Reviews
Manual code review is valuable, but reading your own old code for typos or repeated logic is a time sink. This prompt turns AI into a focused linting partner:
Act as a senior code reviewer. Here is a function/file: [paste code].
Identify only: (1) unused variables, (2) duplicated logic that can be extracted, (3) obvious edge cases that break.
Do not suggest style changes or refactoring beyond what’s listed.
Return a numbered list with the exact line numbers and a one-sentence fix for each.
The magic is the “do not suggest style changes” line. Without it, AI will rewrite your whole file into a different architecture and you’ll spend an hour deciding whether to accept it. With this, you get a checklist you can execute in ten minutes. I run this before every merge request.
The Meeting Notes to Task List Prompt
You’ve just sat through a 45-minute client call. You have messy notes with half-sentences and random decisions. Instead of cleaning them up manually:
Here are my raw notes from a client meeting.
Extract every actionable task and format each as:
- Task name
- Owner (if mentioned, otherwise “me”)
- Deadline (if mentioned, otherwise “no deadline”)
- Dependencies (if any)
Ignore all small talk and context.
Raw notes:
[paste your messy notes]
This turns fifteen minutes of note-cleaning into two minutes. The output is a clean list you can paste directly into your project tracker. If the client mentioned a date as “next Tuesday,” the AI will usually resolve it to the actual date — just double-check.
The Debugging “Last Hour” Prompt
When you’re stuck on a bug for more than an hour, your brain is fried. This prompt forces a structured reset:
I’m stuck on this error: [paste error message or stack trace].
Here is my relevant code: [paste minimal snippet].
Here is what I already tried: [list 2-3 things].
Do not give generic advice like “check your imports.”
Instead, list the three most likely root causes for THIS error in THIS code,
and for each cause, give a specific test line I can run to confirm or deny it.
If you cannot determine a likely cause, say so and ask for one more piece of information.
The key is “what I already tried.” Without it, the AI will repeat your own failed attempts back to you. With it, you force the AI to think narrowly. I’ve solved several “impossible” bugs in under ten minutes this way, simply because the AI suggested a test I hadn’t considered.
The Proposal Boilerplate That Writes Itself
Every freelancer has a standard service — say, building a React dashboard. You still have to write a custom proposal for each lead. Use this:
Write a project proposal for a client who needs [describe project].
Include sections: Overview, My Approach, Timeline, Deliverables, Price Estimate (as a range).
Use a confident but not pushy tone.
Base the technical approach on these keywords: [list 3-5 tech terms].
Do not invent specific features beyond what I list here.
Keep it under 350 words.
I keep a version of this for each service I offer. The AI writes the first draft, I edit the specific numbers and personal touches, and I send it. That’s 30 minutes saved per proposal, and if you send three proposals a week, that’s an hour and a half back.
The “Explain This Legacy Code” Prompt
You inherit a codebase with zero comments. Reading through is painful. Use this prompt to get a map:
Pretend I’m a developer who just joined this project.
Here is a single file: [paste file].
Explain the flow of the program in three sentences.
Then list the main functions and what they do, in the order they are called.
Finally, point out any global state or side effects that could cause bugs later.
Be concise — no praise, no “good code” comments.
This won’t replace reading the code, but it gives you a high-level map in two minutes instead of forty. I use it on every new client’s codebase before my first meeting with them.
These prompts work because they narrow the AI’s scope. They’re not asking for creativity or opinions — they’re asking for extraction, organization, and specific diagnosis. Try one tomorrow. Pick the one that matches your biggest recurring time sink. If it doesn’t save you at least an hour that week, tweak the constraints and try again. That’s the real skill — not prompt engineering, but knowing where your time actually leaks.
I sell these kinds of digital packs in a tiny automated Telegram store — instant USDT delivery. Check it: https://t.me/m3lmhermes_bot
Top comments (0)