Skills / Track anything
Track anything
Set up a low-friction log and a weekly review that makes the data useful. Covers entry structure, capture trigger, honest gaps, summaries, and follow-up. Use when they say "I want to track X", "help me build a habit", or keep abandoning logs.
Use this skill. Nothing to install.
Don't have an agent? · Raw file: skills/track-anything.md
Watch the first two minutes
This is how a tracking session starts with someone whose last five logs all died. Click through it.
Scripted example of a real session.
I always quit by week two
Most tracking dies there, and the skill treats that as a design problem in the tracker rather than a character flaw in you. It builds both halves: a capture flow measured in seconds (two taps to the sheet, dropdowns instead of typing, a trigger moment tied to something that already happens) and a weekly review that pays the data back, because a log nobody reads is a chore with no payoff. It also decides in advance what happens when you miss three days: the gap stays, you resume. If you want a bounded warm-up before tracking something long-term, try a one-week project:
What you end up with
Two tabs: the log you fill in seconds, and the review tab that makes it worth filling. Plus the ritual that connects them.
date exercise weight x reps notes 2026-05-14 bench 140 x 5 2026-05-12 squat 185 x 5 knee ok 2026-05-11 bench 140 x 4 2026-05-09 (blank: missed, gap kept)
sessions this week 3 (last week: 2) current streak 3 weeks logged best bench 145 x 3 (May 2)
- The smallest record that shows progress. Date, the one number that matters, at most two context fields, and an optional notes column. Dropdowns keep labels consistent so the review tab can count them later.
- Computed from the log, never typed. This week vs last week, streak, and best ever: the same shape as simple-dashboard, scaled down to a tab.
- The adjustment is the payoff. Once a week, two minutes, two questions: what does the data say, and what is one change for next week? Averages over weeks beat any single day, so outliers get a shrug.
Questions people actually ask
What happens when I miss three days?
The gap stays blank and you resume. The skill sets this rule on day one, before any streak exists, precisely so a broken streak cannot take the whole log down with it.
Why a spreadsheet instead of a habit app?
Apps shut down, paywall features, and hold your history hostage. The sheet is yours forever, and it feeds directly into chart-the-truth, simple-dashboard, and first-sql when the data grows up.
Can I track several things at once?
The skill will push back. One thing you actually care about, with one number that matters, is what survives. Add a column in month two if a real question demands it.
Can I fill in yesterday from memory?
No. Backfilling manufactures data, and the log stops being evidence the moment you do it. A blank day is honest information about the week.
When does this start paying off?
The first Sunday review, about a week in: you read the week-vs-week numbers and write one adjustment. By then you also have at least seven real entries, which is the skill's definition of done.
Where to go from here
A month of entries banked? The data is ready for more:
Want something concrete to track?
Curious? Read the full skill — the exact instructions your agent gets
--- name: track-anything category: data description: Set up a low-friction log and a weekly review that makes the data useful. Covers entry structure, capture trigger, honest gaps, summaries, and follow-up. Use when they say "I want to track X", "help me build a habit", or keep abandoning logs. --- # track-anything Set up tracking with someone for a thing they actually care about: gym sessions, study hours, money from a side hustle, practice reps, sleep, applications sent. Most tracking stops because logging costs too much and the log is never read. You are building both halves: a capture flow measured in seconds, and a review loop that pays the data back. A spreadsheet is the tool; apps come and go, their sheet is theirs. ## Choose what one entry is Ask what they want to be true in three months, then work backwards to the smallest record that shows progress toward it. One row per event, three to five columns: date, the one number that matters, and at most two context fields. A `notes` column is allowed; required fields beyond date and the number increase the cost of keeping the log. Concrete examples: workout = date, exercise, weight×reps; sales = date, item, amount; study = date, subject, minutes. Resist the urge to track everything — they can add a column in month two if a real question demands it. ## Make logging nearly free Build the sheet together and then attack the capture cost: - Sheet on their phone home screen, sorted newest-first so the entry row is always at the top. - Data validation dropdowns for category fields — tapping beats typing and keeps labels consistent for later counting. - Decide the trigger moment: log immediately after the event, tied to something that already happens (leaving the gym, closing a sale). Avoid a delayed entry plan unless the person's routine supports it. - Missed days stay blank. Backfilling from memory manufactures data; a gap is honest information about the week. Say this now so a broken streak does not end the whole log. ## Build the review loop Tracking without review is a chore with no payoff. Add a `review` tab computed from the log: this week's total or count vs last week, current streak, and best-ever — the same shape as simple-dashboard, scaled down. Then set the ritual: once a week, same day, two minutes. They look at the review tab and answer two questions in a short log line: what does the data say, and what is one adjustment for next week? The adjustment is the point — a log that never changes behavior is a diary, which is fine, but not what they asked for. Averages over weeks beat any single day; teach them to shrug at outliers. ## Done - A sheet with a defined entry format and dropdowns, reachable in two taps - Several historical or current real entries prove the capture flow and formulas - A review tab computes the agreed measure and has a dated first-review reminder - The person knows what to do after a missed entry: leave the gap, resume, and keep the record honest Long-term success is assessed after the first weekly review, not claimed during setup. At that review, record what the data showed and one adjustment. Then: chart-the-truth once a month of data exists, or simple-dashboard if the review tab deserves a URL.