CyberFreedom

Free · no signup

The ADHD AI Prompt Library

By Lachie · CyberFreedom

Every prompt I actually run to get unstuck, free, no email, no signup. Copy one, paste it into whatever AI you already have open (ChatGPT, Claude, whatever), and go. They're organised by the exact moment an ADHD brain stalls. Take what works, ignore the rest.

The bits in [square brackets] are what you fill in. These are starting points, not scripture. Tune them to your voice, but don't polish them. A prompt you run beats a prompt you perfect.

When you can't start

Task initiation. Get a running start you only have to continue.

First three minutes

The core unfreeze.

I have to [task] but I can't get started. The blocker, if I can name it, is [X] (leave this blank if I don't know).

Don't tell me to break it down. I know that doesn't work for me, and choosing how to break it down is its own freeze.

Instead, write me the literal first three minutes of this, as instructions I just follow, present tense, second person. One concrete physical action per minute (minute 1, minute 2, minute 3). Make it boring. Make every action so specific I can't misread it and so small there's no decision left for me to make. Minute 1 must have zero choices in it.

No pep talk, no preamble, no lecture about why I'm stuck. Start at minute 1.
Smallest first deliverable

When the task is too big for "first three minutes".

This task is too big to start: [task].

Don't break it into a plan. Just name the smallest first DELIVERABLE inside it: one concrete thing I could actually finish in 30 minutes, that I could start right now with what's already in front of me. No prerequisites I have to chase, decide, or wait for first.

It has to be something visible or sendable when it's done, not "research", "think about", or "plan".

If "smallest" and "finishable in 30 minutes" pull against each other, pick the one I'm most likely to actually finish.

Give me one deliverable, not three. Then write me the first three minutes of doing THAT, as if you were me: concrete, boring, impossible to misread.

When you can't decide

Decision paralysis. A decision, not more options.

Force me to choose
I'm stuck between these options: [list them plainly].
Here's what actually matters to me, including the petty ones like "least boring to set up" or "fastest to undo if I hate it": [your real criteria].
Don't give me a balanced view. No pros and cons. No "it depends". Commit.
First line, nothing before it: the ONE option you'd pick.
Then three sentences on why, weighed against MY criteria above.
Then one line: the single thing I'm giving up by picking it, versus my next-best option.
If it's honestly a coin-flip, say "coin-flip" and pick whichever is cheapest to reverse.

When you can't stop

Hyperfocus extraction. Pull the director back online.

State-check interrupt

Fire on an interval. Answer in words.

Stop. This is a hyperfocus state check. Don't reassure me or explain anything. Ask me these four questions, one at a time, and make me answer each in one honest line before you ask the next:
1. What am I doing right now, and is it STILL the most valuable use of this time?
2. Did my work actually get better in the last 30 minutes, or am I just polishing?
3. Body check: water, food, movement, bathroom. What have I skipped?
4. If a friend were watching me right now, what would they tell me to do?

Then, based on my answers, give me ONE concrete next action for the next 3 minutes: keep going, or stop and do the thing I'm avoiding (eat, leave for pickup, save and walk away). No lecture. Just the next move.
Capture-and-stop

Make stopping a save point, not a loss.

I have to stop now and I'm scared I'll lose where I am. Make stopping a save point, not a loss.

If you've been with me this session, capture my state from what we've been doing. If you can't see it, ask me these four things in one short batch (I'm in a hurry), then write the note:
1. What was I doing, and why?
2. What's the very next physical action (the first click or line)?
3. What half-finished thought am I holding in my head right now?
4. What was I about to over-polish that I should NOT touch?

Then give me a save point as plain markdown I can paste somewhere, in this exact shape:
- WHAT + WHY: [one line]
- NEXT ACTION (first 3 minutes, concrete): [the literal first step]
- THOUGHT I'M HOLDING: [the fragile idea, before it evaporates]
- DO NOT TOUCH: [the rabbit hole, this is a fence, not a to-do]

Put the NEXT ACTION first and keep it short enough that future-me, flat battery, can be back in the chair in 60 seconds.

When you lost the thread

Working memory. Load the save, don't try to remember.

Session-close handoff

Never just close the laptop.

We're closing this session. Write me a handoff note to future-me, who has a flat battery and no memory of today, so I can resume cold tomorrow. Use what's in this chat; if you can't see today's work, ask me in one line to paste it. If something's missing, write "unknown" rather than inventing it. Give me exactly these, short:

- THE GOAL: the outcome I'm actually after, not the task. One line.
- WHERE I GOT TO: one line.
- NEXT ACTION: the single first thing to do tomorrow, concrete enough to start in 3 minutes without thinking (a file to open, a line to write, a person to message).
- DECISIONS (top 3-5 only): each as "decided X because Y" in one line, so I don't re-litigate them.
- PARKED / STILL OPEN: what I deliberately set aside, so I don't reopen it by accident.

Keep the whole thing under one screen. Output it as plain markdown.
Session-open resume

Load the save; don't try to remember.

It's a new session and I don't want to re-read yesterday.

Find my most recent handoff note (today's if I started one, otherwise yesterday's). If you can't see my files, ask me to paste it in one message and wait. If I don't have a note, don't send me hunting: ask me three quick questions instead (what was I working on, what was the goal, where did I stop) and build the orientation from my answers.

Then orient me in five short lines, no more: where I was, the goal (not just the task), and what I deliberately parked. Don't re-summarise the whole note.

Last, on its own line, give me the one literal next action to take right now, concrete enough that I just do it. Then ask: "Want the first 3 minutes of that?"
Decision log

So you don't re-litigate the same call in three weeks.

Log this decision so future-me can find the reasoning, not just the outcome.

DECISION: [what I decided]
WHY: [one line]
INSTEAD OF: [the alternatives I rejected, and the one-line reason each lost]

Format it as a single dated, append-ready entry I can paste into my decision log (add today's date and a 3-word title at the top so I can scan for it later). Only use what I gave you; if a field is thin, leave it short rather than inventing reasons I didn't have.

When it's the wrong task for the moment

Energy and state. Best-FIT task, not most-important.

Energy-aware router
Here's my current state, honestly: [energy level, focus, mood, time available, and anything physical: pain, hunger, restlessness, meds]. If that's too much to fill in, just tell me roughly.

Here's my task pool: [paste the list]. I probably haven't tagged these by the energy they need, so work out the state-cost of each yourself.

Tell me the ONE task to do right now that best FITS the state I'm actually in, not the most important task. One line on why it fits.

If nothing in the pool fits my state, say so. And whether or not a task fits, if my state clearly needs it (dehydrated, hungry, in pain), tell me the state-change action to do first: food, water, a walk, or a short break.

When you can't ship

Rejection sensitivity. Let the AI take the first hit.

Exposure-absorber draft
I need to [send / post / publish] this but I'm frozen on the exposure of it.
What it needs to say, however rough: [dump the points, bullets are fine].
Draft it for me, in my voice, ready to go out. Make it good enough that my job is to tweak and send, not to write. Give me one clean draft, not three.
Keep it [honest / warm / brief], and don't make it grovel or over-explain. That's the RSD talking, not good writing.
Harsh-but-kind pre-read
Before I ship this, be the toughest fair reader it'll meet.
Here's the draft: [paste it].
What will the harshest reasonable person say? Give me the 3 real weaknesses, ranked, and the fastest fix for each.
Then one line: whether this is shippable once those fixes are in.
Be direct, but don't catastrophise, and don't rewrite the whole thing. I'd rather hear it from you now than imagine it from them later.
Reply playbook
I'm about to publish [thing].
Draft me a short reply playbook for the reactions I'm dreading: a harsh-but-fair criticism, a correction I'll have to accept, and total silence.
Keep my replies brief, non-grovelling, and human.
I want to react from something I wrote while calm, not something I improvise while flinching.

The daily loop

Three prompts that run your day: open, midday, close.

Morning open
Good morning.
Find my most recent handoff note (yesterday's, or today's if I started one). If you can't see my files, ask me to paste it. If I don't have one, ask me three quick questions instead of sending me hunting.
1. Orient me in 5 lines: where I was, the goal, the literal next action.
2. Here's my honest state right now: [energy / focus / time / body]. Here's my task pool: [paste]. Route me to the best-FIT first task.
3. For that task, write me the first 3 minutes as if you were me: concrete, boring, impossible to misread.
Midday check
Midday check. Ask me these one at a time and wait for each answer before the next:
1. What am I doing, and is it still the best fit for my state right now?
2. Body: water, food, movement, bathroom. What have I skipped?
3. Has my battery moved since I picked this task?

Then give me one instruction: keep going, switch tasks, ship, park, or take a body reset. If you need my task pool to re-route me, ask me to paste it. If I'm polishing, tell me to stop and ship or park it.
Session close
Closing the session.
1. Write the handoff for future-me: goal, next 3 minutes, decisions and why, what's parked. Keep it under one screen, plain markdown.
2. What's the ONE finished thing I can ship before I stop? It has to take 10 minutes or less. If it already exists, critique only what's blocking me from sending it; if it doesn't, draft the smallest sendable version. End with the exact next action.

The one question that protects the whole system

Run it daily. It's the only metric the traps can't fool.

Did anything ship?
Did anything SHIP today, or did I just build / engage?
One honest line.
If nothing shipped, name one tiny thing I can ship in the next 10 minutes. No pep talk.

If any of these just worked, read this before you close the tab. These prompts are the what. On their own they'll get you unstuck plenty of times. But a prompt you don't understand is one you'll drop the first week it doesn't land, which is exactly how every productivity system you've tried died: it worked until it didn't, you had no idea why, so you blamed yourself and walked away.

The manual is the why. The same six failure modes, but each mapped to the wiring underneath it, the back-and-forth that actually gets the good answer (not just the one prompt), and a real example from my own work for each. Plus the chapter most books skip, when AI makes ADHD worse, and a 30-day install plan so you bolt on one workflow at a time instead of trying all six and dropping the lot. There's a fill-in workbook too, where you build your own one-page operating manual for the wall.

I wrote it using these exact prompts, after three decades of being sure the right system would finally fix me. None did. This was the first thing that routed around the wiring instead of fighting it. Pay what you want from $9, and if money's tight the floor is the price, no story needed.