Your athletic record
A readable index of your training, with the detail close at hand. Local files, open structure, explicit sources.
Use your history
Your watch has the session. Your workout log has the score. Your notes have the context. Bringing them together lets an agent work across those records instead of answering from one app’s view.
These are questions to ask an agent with the ATH skill and access to your folder. Its answer depends on what you have imported and logged.
Rebuild your history
A year of device sessions might say when you trained, for how long and at what heart rate. A spreadsheet has your sets and scores. A folder of PDFs has the programmed workouts. Give your agent those files, plus any notebook photos or messages you want included, and ask it to assemble a dated training history.
The agent can organise the workout details, propose matches to imported sessions and help save the results and links in ATH. A generic strength session can then sit alongside the squats you actually did and the weights you logged. Review ambiguous matches, duplicate entries and planned workouts without a recorded result before saving; a programme alone does not prove you completed it.
Ask your agenthere are my device exports and workout logs from last year. help me put them together and link the workout details to the right sessions. show me gaps, duplicates and anything you need me to confirm
Switch devices
Moving from an Apple Watch to WHOOP should not mean your training history starts again. Import both exports and keep your workout results alongside them. Your agent can review a training block that spans the switch, or find the sessions leading up to a personal best recorded with your old device.
If you wore both devices for a while, ask the agent to identify overlapping sessions so a weekly review does not count the same workout twice. Keep each device’s original measurements and baselines separate; matching sessions does not make their scores equivalent.
Ask your agenti switched from apple watch to whoop in june. put the full year of training together and flag sessions both devices recorded so we don’t count them twice
Compare attempts
Compare a Fran result from your notebook with the wearable session it belongs to, then look at the training and sleep leading into each attempt. You can investigate a slower time without manually matching dates across apps.
Ask your agentcompare my last three fran attempts with the week before each one. what changed and what data is missing
Connect different logs
The same workout can be filed under a benchmark name, a date, a coach’s shorthand or a generic activity label. An agent can look through the descriptions, movements and recorded scores to suggest comparable attempts, even when a search for the workout’s name would miss them.
Use those matches to recover an old best, compare a repeated training block or see how lifting progressed during run-heavy weeks. Changes in load, distance or scaling matter: keep them visible rather than treating every similar session as the same test.
Ask your agentfind workouts like today’s in my old logs even if i called them something different. compare the weights, reps and times where they’re recorded and show me any matches you’re unsure about
Review a training block
Bring together logged scores, imported sessions and notes about how training felt. Ask for a short review grounded in actual attempts, then choose what to share with your coach. This combines context that might otherwise be spread across a wearable app and a notebook.
Ask your agentsummarize my last six weeks for my coach. include benchmark changes, training consistency and my notes. flag weeks with missing data
Build custom views
Ask your agent to make a chart or a small local dashboard from your record. For example, put repeated workout scores beside the previous week’s session count and sleep history, with links back to the underlying entries. The agent builds the view; ATH provides the data.
Ask your agentmake a chart of my fran times with the previous week’s training and sleep. show gaps instead of filling them in
Compare models
Give different models the same historical evidence and compare their predictions with results you already recorded. ATH’s backtest uses the evidence available before each attempt, writes a separate report and leaves your record unchanged. This makes model choice something you can evaluate against your own training.
Ask your agentcompare two models on my past results. show where each was wrong and how many attempts the comparison covers
File structure
athlete.ath.json holds your athlete details, sources, benchmarks, signals, results and predictions. High-frequency samples are kept in series/ so the main document stays readable.
$ath-mike/$ athlete.ath.json$ series/Data sources
A device measurement identifies the source that produced it. An observation you report about yourself stays separate. A number typed by hand is marked as manual; it does not become a wearable reading.
ATH keeps averages separate by source. Two devices measuring the same quantity do not silently become one baseline.
Benchmarks
Benchmarks describe a repeatable effort scored by time, repetitions or load. Fran, a 5k, a row and a strength test can all have a history. Segments keep the parts of a longer effort visible.
Validation
Validation checks the structure and references, including missing sample files. Run it after editing the record or moving a backup.
$ath check