| name | weekly-load-report |
| description | Produces the Northfield FC one-page weekly training-load PDF report for the head coach. Aggregates cleaned GPS data per player across the training week, compares to the previous week, flags load spikes, and renders a standardised Typst PDF. |
Weekly training-load report
When to use this skill
Use this skill when the user asks for the "weekly load report", "coach's weekly report", "training-load PDF", or anything that sounds like they want this week's training-load summary for the head coach. Default squad is U23 unless the user says otherwise.
Inputs
data/gps_training_clean.csv - the current week, cleaned by the Module 4 tool. If this file does not exist, run the cleaner first (rules in ../../module-4-build-a-tool/README.md) or clean inline using the same rules.data/gps_training_week_prior.csv - the previous week, tidy by construction.data/squad_roster.csv - player-to-squad/position map.
If the user passes a different file path for the current week, use that instead.
What the skill produces
A one-page PDF at the path the user specified (default weekly_report_<iso-date-of-monday>.pdf in the current working directory).
The PDF contains, in order:
- Header: "Northfield FC - Weekly Training Load" and the week range (e.g. "20-26 April 2026").
- Top-line paragraph: 2-3 sentences summarising total squad distance, number of sessions, and how the week compares to last week at squad level.
- Player table: one row per player in the chosen squad, sorted by total_distance_m descending. Columns: player_id, position, sessions, total_distance_km, high_speed_running_m, sprint_distance_m, delta_vs_prev_pct. Rows coloured:
- red if
|delta_vs_prev_pct| > 20
- amber if
15 < |delta_vs_prev_pct| <= 20
- uncoloured otherwise
- Footer: generated timestamp and the phrase "Synthetic data - teaching use only".
Aggregation rules
- Week runs Monday to Sunday.
- Only include the current squad (default U23).
total_distance_km = sum of total_distance_m across the week, divided by 1000, rounded to 1 decimal.high_speed_running_m and sprint_distance_m = sums across the week, rounded to 0.delta_vs_prev_pct = (current_total_distance - prev_total_distance) / prev_total_distance * 100, rounded to 1 decimal. If a player has no previous-week data, display - and do not flag.
How to fill the template
The Typst template lives at template.typ next to this SKILL.md. It expects a JSON blob at report_data.json (same folder) with this shape:
{
"week_label": "20-26 April 2026",
"squad": "U23",
"summary_paragraph": "Two pitch sessions this week...",
"rows": [
{"player_id": "NF001", "position": "CB", "sessions": 2, "total_distance_km": 14.2, "hsr_m": 980, "sprint_m": 320, "delta_pct": 3.1, "flag": "none"},
...
],
"generated_at": "2026-04-26T19:41:00+10:00"
}
Write report_data.json in this skill folder, then compile with:
typst compile skills/weekly-load-report/template.typ <output-pdf-path>
Typst resolves report_data.json relative to the template file; do not change that path.
Post-run summary
After producing the PDF, print a short terminal summary:
✅ Wrote: <path>
Squad: <squad>
Week: <week_label>
Players included: <n>
Flagged red: <n>
Flagged amber: <n>
If no red or amber flags, say "No load flags this week." explicitly - silence looks like a bug.
Safety
- Never include player real names, only
NF### IDs. (This repo has no real names, but the rule stands for when you adapt this to a real club.)
- Never write the PDF to a tracked location. Keep it inside the relevant module folder or
outputs/.
