Most planning systems die in the gap between two and three. People build the tracker, feel productive, and never review it, so nothing ever adapts. The review is the part that compounds, and it is why the setup below spends more time on it than on the plan.
You are my accountability coach for the goal in this project. Your job is to keep me moving, not to make me feel good about a plan I am not following.
Always work from the plan and the tracker in this project rather than from memory, and ask for the current numbers if you do not have them. When I report a week, respond with one honest read of what the data shows, then exactly one adjustment for the coming week, never a list of five improvements, because a list is how I end up changing nothing. Count effort in sessions completed, never in whether the week was perfect, and never suggest I start again from scratch.
Be warm and direct at the same time. Do not lecture me about discipline, do not use motivational filler, and if I have missed the same thing three weeks running, say so plainly and tell me the plan is wrong rather than telling me to try harder. Keep every response short enough that I actually read it on a bad day.Run this first, inside the project. The output is the plan every later prompt reads from. The part people get wrong is the schedule: give it the week you actually have, including the evenings you know you will be useless.
I want to build a plan for this goal: [the goal, stated as something finished, for example "pass the four exams in my semester" or "publish twelve reels" or "ship the first version of my app"]. My hard deadline is [date]. Here is my honest weekly availability, not my aspirational one: [days and rough hours, including the times you know you will not work].
Work backwards from the deadline. Break the goal into the specific pieces of work that have to exist by the end, put them in the order they actually depend on each other, and tell me which piece is the one that quietly determines whether the whole thing lands. Then convert that into a weekly session count: how many working sessions per week, how long each one is, and what a single session is supposed to produce. Define a session as something I can finish in one sitting and tick off, never as a vague block of time.
Two rules while you do this. Assume I will lose roughly one week in every four to life, and build that slack in rather than planning for a perfect run. And if the deadline and my real availability do not fit, say so directly in the first line and tell me what has to shrink: the scope, the quality bar, or the deadline. Do not quietly compress the plan to make the numbers work. Finish by asking me up to three questions if anything important is missing.The unit matters more than the tool. A tracker that measures "was today good" gives you a broken streak and a reason to quit. A tracker that counts completed sessions against a weekly target survives a terrible week, because a terrible week is a three out of five, not a failure.
Build me a progress tracker for the plan above, as a single self-contained HTML file I can save and open in my browser, with the data stored in the file so I can edit it in a text editor.
It needs: a header showing my goal, the deadline, and how many weeks are left. A row per week from now until the deadline, with columns for sessions targeted, sessions completed, what was produced that week in a few words, and a one line note. A running total of sessions completed against sessions targeted for the whole goal. And a simple visual bar showing progress towards the total, so I can see the shape of it at a glance.
Design rules. It has to count completed sessions, never streaks and never day-by-day perfection, because a broken streak is the thing that makes people quit. Show the running total as "42 of 96 sessions" rather than a percentage of days survived. Keep it plain and readable rather than decorative, and add a short comment at the top of the file explaining exactly how to update it each week.
Also give me a markdown version of the same table, so I can paste it back to you in our weekly reviews.This is the part most people skip, and it is the part that turns a plan into a system. Fifteen minutes, same time every week, and you leave with exactly one thing changed.
Weekly review. Here is what actually happened: I targeted [X] sessions and completed [Y]. Here is what I produced: [a few words per session, or what got finished]. Here is what got in the way, honestly: [the real reasons, including the boring ones].
Do four things, in this order and briefly. One, tell me where I am against the plan in a single sentence, using the session numbers rather than a judgement about my effort. Two, name the pattern you can see across the weeks so far, especially if the same day, time or type of session keeps failing, and say if it is too early to call a pattern. Three, give me exactly one adjustment for next week, and make it a change to the plan or the schedule rather than a change to my willpower. Four, tell me my session target for next week and what the first session should produce.
Do not give me a list of improvements, do not praise me for showing up, and if my completed sessions have been under target three weeks running, tell me the target is wrong and propose a lower one I would actually hit.Everyone drops a few days. The damage is almost never the lost sessions, it is the restart: throwing out the plan, rebuilding it from scratch on a Sunday, and treating four days off as a reason to start again in January. This prompt exists so that never happens.
I missed [number] days or sessions. Do not rebuild the plan from scratch and do not reset anything, re-fit what is left.
Take the sessions I still have to complete and the days I actually have left before the deadline, and tell me the new weekly session count. If that number is now higher than what I have ever actually hit in a week, say so plainly and give me two options instead of a fantasy schedule: what to cut from the scope, or what a realistic new deadline looks like. Be specific about what gets cut, name the actual piece of work, do not tell me to "reduce scope".
Then give me one small change that makes the next three days easier to start, for example a shorter first session or a different time of day. Keep it to a few lines and skip the encouragement, I want the maths and the next action.Once the weekly review has run a few times, stop pasting the prompt. A skill is a small folder with a SKILL.md inside holding a name, a description of when to use it, and your instructions. Claude pulls it in when your request matches, so the review becomes one word. Skills work on every plan, Free through Enterprise, and there is no code involved.
Build me a reusable Claude skill called "checkin" that runs my weekly accountability review.
Use skill-creator to write it. The SKILL.md description should say to use it whenever I type /checkin or say it is time for my weekly review. The instructions should say: first ask me for three things and nothing else, my target sessions, my completed sessions, and what actually got in the way. Then respond in four short parts, where I am against the plan in one sentence using session numbers, the pattern across the weeks so far, exactly one adjustment to the plan or schedule rather than to my willpower, and next week's target with what the first session should produce.
Add these standing rules to the skill: never suggest starting over, never give a list of improvements, never open with praise, and if completed sessions have been under target three weeks running, say the target is wrong and propose a lower one. Keep every response short. Then package it as a zipped folder I can upload.~/.claude/skills/ instead. Anthropic's creating custom skills article has the same steps.Claude cannot see your screen, cannot verify a session happened, and will not chase you. What it does well is hold the whole plan in one place, notice patterns in your own reported numbers faster than you will, and give you a re-plan in thirty seconds so a missed week never turns into a quit.
Harkin et al., Does monitoring goal progress promote goal attainment? Psychological Bulletin, 2016 (138 studies, N = 19,951)
Creating and managing projects, including the five project limit on free accounts
Understanding Claude's personalization features, on profile and project instructions
Anthropic on artifacts, including plan availability and downloading files