agentleFS
totorototototorototo/ledger/.claude/skills/load-check/SKILL.md

load-check

Skill0 starsChanged 12 days agoMITView on GitHub

Check the runner's recent training load for warning signs — load-ratio spikes, sudden jumps in weekly volume, easy vs hard balance — and say whether it looks like overreaching. Use when asked "am I overdoing it", "is my load OK", about injury risk, ramp rate, or acute:chronic ratio.

  • Reads credentials

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 0 are distinctive, and 8 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 8

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Every number comes from data/metrics.json, written by npm run export from the same sums the app shows (src/utils/insights.js).

    line 8
  2. Very specific:

    Read data/metrics.json. If it's missing, its until isn't today, or its version isn't 1, run npm run export (it needs …

    line 14
  3. Very specific:

    If empty is true, say there are no activities and stop.

    line 17
  4. Very specific:

    Check coverage.withLoad against coverage.kept: if many activities have no load, the ratio understates the real load: say so.

    line 47
  5. Very specific:

    Read each flag with the week's notes: an abrupt drop on a planned or health week is explained, so say what explains it instead of asking.

    line 78
  6. Very specific:

    After the write-up, the same reading as a page in Ledger's style (scripts/reports/load-check.html).

    line 134
  7. Very specific:

    Write data/reports/load-check.notes.json (shape below).

    line 137
  8. Very specific:

    Run npm run report -- load-check.

    line 139

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Load check

Every number comes from data/metrics.json, written by npm run export from the same sums the app shows (src/utils/insights.js). Never compute a figure yourself: quote it, or say it isn't there. Your job is to interpret.

1. Get fresh numbers

  1. Read data/metrics.json. If it's missing, its until isn't today, or its version isn't 1, run npm run export (it needs GARMIN_PERSONAL_API_KEY in .env; if that fails, say so and stop, never guess numbers).
  2. If empty is true, say there are no activities and stop.

2. Read

The last 8 entries of weeks[] (Monday-start weeks, the last one ends at until and may be partial), plus lastWeek.load, lastWeek.hrv, lastWeek.restingHr, lastWeek.sleep and lastWeek.notes for the rolling 7 days to today:

FieldMeaning
weekthe Monday it starts
km, hours, climbM, runKm, loadthe week's totals (every sport; runKm running only)
kmChangekm against the week before, "+35%"
acute, chronic, ratioat the week's end: 7-day load, 28-day load per week, their ratio
intensityeasyPct / moderatePct / hardPct % of heart-rate time (Z1–2 / Z3 / Z4–5)
partialtrue when the week ends after until: its totals are of the days so far
hrvnights, average (ms), belowBand / aboveBand: nights outside the balanced band of the runner's own baseline (Garmin's); null with no reading
effectssessions per Garmin training-effect label (TEMPO, RECOVERY…)
restingHrdays, average (bpm), baseline (median of the 28 days before), raised: days ≥ 5 bpm over it; null with no reading
sleepnights, averageHours, averageScore (Garmin's 0–100), under7h: nights under 7 h; null with no reading
notesthe runner's plan entries overlapping the week, "planned: recovery week", "health: cold"

lastWeek.hrv.days[] has each night: date, lastNightAvg, status (Garmin's: BALANCED, UNBALANCED, LOW…), balancedLow, balancedUpper. coverage.hrvFirst is the first night with a reading: before it, hrv is null and HRV says nothing. Same for restingHrFirst and sleepFirst.

plan (null without data/plan.md): nextRace (date, text, daysTo), races[], and ignored, the plan's lines the export couldn't read: name them so the runner can fix them.

Check coverage.withLoad against coverage.kept: if many activities have no load, the ratio understates the real load: say so.

3. Flags

Raise a flag for each that holds, quoting the week and the numbers:

FlagRuleSource
Load spikeratio > 1.5Gabbett, Br J Sports Med 2016
Sustained high ratioratio > 1.3 two weeks runningsame; the "sweet spot" is about 0.8–1.3
Volume jumpkmChange > +30% on a week of more than 2 hNielsen et al., J Orthop Sports Phys Ther 2014: novice runners who raised weekly distance more than 30% were injured more often
Grey-zone driftintensity.moderatePct > 30% for 3+ weeksSeiler 2010: ~80% easy is the norm in endurance sport
Abrupt dropratio < 0.6not a risk in itself: ask whether it was planned (illness, injury, taper)
HRV under bandhrv.belowBand ≥ 3 nights in a weekPlews et al., Sports Med 2013: HRV falling under the athlete's own baseline is the sign the load isn't being absorbed
Resting HR raisedrestingHr.raised ≥ 3 days in a weeka supporting sign only: resting HR responds to fatigue less reliably than HRV (Buchheit, Front Physiol 2014)
Short sleepsleep.under7h ≥ 3 nights in a weekunder 7 h is below the adult range (Hirshkowitz et al., Sleep Health 2015); a reason for HRV under band, not a load flag

The "10% rule" has little evidence behind it (Buist et al., Am J Sports Med 2008): don't cite it as a rule.

A week with partial: true has low totals and kmChange by construction: never flag its volume, say it's partial. Its ratio is still a whole rolling 7 days and counts.

Split the flags in two:

  • Now: the rolling 7 days (lastWeek.load) and the last two weeks of weeks[] that aren't partial. These decide the verdict.
  • Earlier: flags 3 to 8 weeks back. Context only: they never set the verdict, but say how they lead to now ("the drop follows September's spike").

Read each flag with the week's notes: an abrupt drop on a planned or health week is explained, so say what explains it instead of asking; a spike in a race week is the race. With no plan, ask as below. Short sleep beside HRV under band on a light week points to sleep, not training: say so.

4. Verdict

About now only. A rising flag is a load spike, a sustained high ratio or a volume jump. Take the first that holds, so exactly one always applies:

  1. Overreaching risk: a rising flag now and HRV under band now; or, with no HRV, a rising flag now plus resting HR raised now, or a load spike now plus a volume jump or a sustained high ratio now. Resting HR raised beside HRV under band makes it firmer: say so.
  2. Watch: any rising flag now, lastWeek.load.ratio over 1.3, or HRV under band now without rising load (illness, stress, sleep: say which the notes and sleep point to, else ask).
  3. Lighter: lastWeek.load.ratio under 0.8. Not a warning: ask whether it was planned.
  4. Steady: everything else: lastWeek.load.ratio 0.8–1.3, with no rising flag and HRV not under band.

Abrupt drop, grey-zone drift, short sleep and resting HR raised on its own never set the verdict: list them under Now beside it, so a Steady week can carry a flag. A rising flag with HRV in its band says the load is being absorbed so far: say so, it's the most useful line here.

Be honest about what the data can't show: soreness and how training feels, unless the plan's health notes say it, and no HRV, resting HR or sleep before their coverage first dates. Overreaching is felt before it shows in load: invite the runner to weigh how they feel.

5. Write

**Load check, <until>** — <Steady | Lighter | Watch | Overreaching risk>

Load ratio now <lastWeek.load.ratio> (7-day <acute> / 28-day <chronic> per week)
HRV last 7 nights: <average> ms, <belowBand> under / <aboveBand> over band (or "no reading")
Resting HR <average> bpm against <baseline>, <raised> days raised · Sleep <averageHours> h, <under7h> nights under 7 h
Next race: <text> in <daysTo> days (only with a plan)

| Week | km | h | Load | Ratio | Δ km | Easy % | HRV (under band) | Sleep < 7 h | Notes |
|------|----|---|------|-------|------|--------|------------------|-------------|-------|
(last 8 weeks, oldest first; a partial week marked "(so far)")

**Now**
- <flag>: <week, numbers, why it matters> (or "No flags.")

**Earlier** (only when there is one)
- <flag>: <week, numbers>, and how it leads to now

**What to do with it**
<one or two sentences: hold volume, add an easy week, nothing to change>

End with: Based on Garmin's activity and wellness summaries; a warning light, not a diagnosis, and not medical advice.

6. Publish the page

After the write-up, the same reading as a page in Ledger's style (scripts/reports/load-check.html). The page takes its numbers from metrics.json itself: you write only words.

  1. Write data/reports/load-check.notes.json (shape below). Text may quote numbers from metrics.json, as the write-up does; never one of your own.
  2. Run npm run report -- load-check. It checks the notes, writes data/reports/load-check.html and prints the page's link, if it has one. On an error, fix the notes it names and run it again.
  3. Publish data/reports/load-check.html with the Artifact tool. With a link: read that artifact first, then publish with it as url, so the page updates in place. Without one: publish it new (icon chart), then record its link with npm run report -- load-check --url <link>.
  4. End the reply with the link. The page uploads the account's data to claude.ai, private until the runner shares it: never share it yourself.
{
  "verdict": "Steady | Lighter | Watch | Overreaching risk",
  "takeaway": "one sentence: the verdict and why",
  "now": [{ "flag": "Load spike", "text": "week, numbers, why it matters" }],
  "earlier": [{ "flag": "…", "text": "…, and how it leads to now" }],
  "action": "the What to do with it sentence"
}

now empty shows "No flags".

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.