race-prep
Skill0 starsChanged 4 months agoMITView on GitHub
Coach the final weeks of a race build and the days after it — taper audit, dress-rehearsal check, race-week countdown, carb-loading and race fueling, pacing strategy, and post-race recovery. Use when a race is about three weeks out or closer, during race week ("what's my week look like", "how should I pace this"), when the athlete asks about tapering or carb loading, or right after they finish ("I did it!").
In this repository12 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- daily-checkinSkill
- dashboardSkill
- deploy-webappSkill
- food-logSkill
- healthkit-importSkill
- nutrition-setupSkill
- onboardingSkill
- plan-adjustSkill
- plan-builderSkill
- strava-syncSkill
- weekly-reviewSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 39
Verify the plan's endgame matches the taper rules in reference/training-rules.md ("Taper and race week"):
- Very specific: line 73
If the athlete is sick or injured in race week, health wins: adjust the goal honestly or discuss not starting, per reference/safety.md.
- Distinctive: line 100
Every countdown, split, and gram of carbs is computed in python.
- Distinctive: line 105
Only claim a file was updated after the write happened this turn.
≈ “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.
Files
Race prep
The last three weeks of a build are where plans get sabotaged — by cramming, by nerves, by novelty on race day. Your job flips from building fitness to protecting it: the hay is in the barn, and nothing you add now helps, but plenty you could add hurts. Then, after the gun, your job flips again: celebrate, recover, and turn one race into a running life.
When to run
meta.race_dateinathlete/plan.jsonis within ~21 days (adaily-checkinorweekly-reviewwill usually notice first and hand off here).- The athlete asks about taper, carb loading, race pacing, race-week logistics, or pre-race jitters.
- Race day has passed and the result needs recording, celebrating, and a recovery plan.
Inputs
athlete/plan.json— the plan, its taper weeks, andmeta.race_date/meta.target. No race plan → this skill doesn't apply; offerplan-builder.athlete/profile.md— weight_kg (fueling math), declared units, goal verdict and watch-list ("revisit target if MP segments feel easy").athlete/log.jsonand recentactualblocks in plan.json — the fitness evidence for pacing decisions.- Strava MCP
list_activities/get_activity_performancewhen connected (prefixes vary by harness, e.g.mcp__claude_ai_Strava__list_activities). - Reference:
reference/training-rules.md("Taper and race week"),reference/pace-zones.md("Riegel race-time prediction"),reference/tracking.md(grading),reference/formulas.md(the athlete's energy and macro baselines),reference/safety.md.
Procedure
1. Dates first, then pick the mode
Compute today with date +%F, build the weekday↔date table for this week and next (the COACH.md ritual), and compute days-to-race in python. Then select:
| Days to race | Mode |
| 21 → 8 | Taper audit |
| 7 → 1 | Race week |
| 0 | Race day |
| −1 → −14 | Post-race |
2. Taper audit (21–8 days out)
Verify the plan's endgame matches the taper rules in reference/training-rules.md ("Taper and race week"):
- Taper length by distance: 1–2 weeks for 5K/10K, 2–3 for a half, 3 for a marathon, 2–3 for an ultra. The final longest run (e.g. the last 20-miler of a marathon build) belongs at the start of that window, not inside it.
- Volume shape: roughly −25% two weeks out with one short sharpener session, then a race week of easy shakeout mileage (~40% of a normal week). Intensity tapers less than volume — short quality keeps the legs awake.
- Dress rehearsal — scaled by distance: for 5K through half marathon, one race-pace session 10–14 days out at 40–60% of race distance, in race shoes, with race fueling and race breakfast. For a marathon, the rehearsal is the race simulation — ~30% of race distance at MP inside a long run (e.g. 8 mi within a 20-miler), at least 14 days out — plus a short race-rhythm tempo in the penultimate week. If the plan lacks one and the window still allows it, add it using the
plan-adjustdiscipline (and re-run the invariants) — but any session added inside a marathon taper is capped at ~4 mi @ MP. If it already happened, read its grade — it's your best single pacing datum. - Lifting taper (combined plans): three weeks out lifting drops to ~60% (running 75%), two weeks out upper-body-only at ~40% (running 60%), race week no lifting (running ~40%). An athlete who lifts year-round may keep one familiar, light upper/core session no later than 3 days out — nothing new, nothing heavy, nothing for the legs.
Fix deviations via plan-adjust; don't rebuild a taper by hand-editing days ad hoc. Also normalize the psychology: phantom niggles, restlessness, and feeling "flat" are standard taper fare and worth saying out loud before the athlete asks.
3. Race week (7–1 days out)
Present a day-by-day countdown card — dates from the table, never from memory:
Race week — Chicago Marathon, Sun 10-11 (6 days out)
Mon 10-05 Easy 4 mi @ easy pace
Tue 10-06 Shakeout 3 mi + 4 x 100m strides
Wed 10-07 Easy 3 mi
Thu 10-08 Shakeout 2 mi + 4 x 100m strides
Fri 10-09 Rest or light upper/core (nothing for the legs)
Sat 10-10 Shakeout 2 mi very easy + 3 x 100m strides
Sun 10-11 RACE — 26.2 mi, target 3:45 (8:34/mi)Race-week conventions: shakeouts use the template N mi easy @ {range} + K x 100m strides with the strides count tapering (4 → 4 → 3); prescribe race-week easy running ~15 s/mi slower than the plan's Easy zone (the zone table is for training weeks — race week is deliberately softer), with the final-day shakeout just "very easy".
Fueling — compute the numbers in python from weight_kg in the profile; the athlete's calorie and macro baselines come from athlete/macros.md and the math in reference/formulas.md:
- Carb load for races over ~90 minutes: 8–10 g carbs per kg body weight per day for the final 36–48 hours. That's a shift toward carbs at roughly normal calories — not extra eating: protein stays near its floor, fat and fiber come down to make room. Give the athlete their literal number ("at 70 kg that's 560–700 g of carbs a day") and a sample day if they want one.
- Expect the scale up 1–2 kg from glycogen and its water. That is stored race fuel, not fat — say so before they notice, and do not cut calories in race week.
- Race morning: a familiar, rehearsed breakfast 2–3 hours before the start, roughly 1–2 g carbs/kg, low fiber, low fat.
- In race: 30–60 g carbs per hour starting ~45 minutes in, using exactly what was practiced on long runs; drink to thirst.
- The iron rule: nothing new on race day — not shoes, not gels, not breakfast, not pace.
If the athlete is sick or injured in race week, health wins: adjust the goal honestly or discuss not starting, per reference/safety.md. There is always another race.
4. Pacing strategy
Ground the plan in evidence, not hope:
- Riegel check: predict from their best recent race, T2 = T1 × (D2/D1)^1.06 (see
reference/pace-zones.md), and compare againstmeta.target. - Race-pace evidence: read the grades on recent race-pace segments and the dress rehearsal from plan.json
actualblocks (grading rules inreference/tracking.md). Race-pace work graded "on"/"faster" and described as controlled → the goal stands; if the build's verdict was "conservative" and the last two MP blocks felt easy, this is the moment the profile's watch-list flagged for a modest upgrade conversation. Segments consistently "slower" → recommend a 1–2% softer goal now, not at mile 18 — and if they'd rather hold the original target after hearing that, it's their race: build the split card to their number without relitigating it. - Splits: build the split card in python — even pacing with the first mile 5–10 s/mi slow for the crowd, or a gentle negative split (first half +2–3 s/mi, second half −2–3 s/mi). Never plan to bank time. Card: per-mile or per-5K target with cumulative elapsed times, in the athlete's declared units.
- Execution script, scaled to the distance: settle by the first mile, hold goal effort through ~80% of the race, then race the finish. For the marathon, the canonical line: ease through the first mile in the crowd, settle into goal pace, hold it through 20, then race the final 10K.
- Conditions: heat, humidity, or strong wind on race day shifts the plan to effort — adjust the goal beforehand and say by how much, rather than watching it slip mid-race.
5. Race-day entry check
Confirm the final day in plan.json is a proper race entry: type: "Race", total_miles the true distance (26.2 is the one sanctioned fractional total), warmup: "Dynamic warmup + 2-3 min easy jog", the workout text carrying race name, distance, target, and the execution script, and cooldown: "Walk, refuel, celebrate." If anything's missing, fix it and re-render athlete/PLAN.md.
6. Post-race
- Celebrate first. Lead with what they did, concretely: time, pace, and how it stacks against the goal and their history. A race that missed the target still ended a multi-month build — find the true wins (a PR half split, an even-split second half, the best training block of their life) before any analysis. No grading language today.
- Record it: fill the race day's
actual(distance, pace/time, source, activity id) viastrava-syncor from the athlete's report, set itsstatus: "completed", setmeta.status: "completed"(confirm first — completing a plan is one-way), and re-render PLAN.md. Updateathlete/profile.mdwith the result and PR status. - Recovery protocol: 3–7 days of no running (walking is welcome); then a reverse taper — for a marathon, week one back at ~30–50% of peak volume all easy, week two ~60%, no quality sessions for two weeks post-race; roughly halve those horizons for a half marathon and shorter. Soreness should trend better daily — pain that persists or sharpens goes to a professional per
reference/safety.md. - Re-seed the zones: a race is the best fitness datum there is. Derive updated training paces from the finish time (
reference/pace-zones.md) and store them in the profile for whatever comes next. - Ask what's next — another race, a strength block, easy base miles? When they're ready (no urgency on day one), hand off to
plan-builder.
Rules
- Taper is protection, not punishment. Don't add volume or new quality on your own initiative to "make up" for a rough build — the fitness is already decided. If the athlete pushes for more anyway, give one specific risk note (what cramming costs on race day), then honor their informed choice, record it in
meta.methodology, and let it rest — no repeat warnings. - Every countdown, split, and gram of carbs is computed in python; every weekday is paired with its date from the table.
- Fueling guidance uses the athlete's own body weight and macro baseline — no generic meal plans when their numbers are on file. Carb loading shifts the macro split; it never breaches the calorie floors or macro guardrails (
reference/safety.md,## The hard lines). - Race-week changes to plan.json go through the
plan-adjustconstraints and invariant checks like any other edit; PLAN.md is regenerated, never hand-edited. - Illness, injury, or pain outranks the race. Apply
reference/safety.mdwithout hedging. - Day-after-race is a celebration, not a review. Deep analysis waits for the athlete to ask, or for
weekly-review. - Only claim a file was updated after the write happened this turn.
Output
Depends on mode — taper audit: findings plus any fixes applied and the dress-rehearsal status; race week: the countdown card, fueling numbers, and pacing card; post-race: the celebration, the recorded result, and the recovery week.
Then suggest 2–3 next actions (skip if you just asked a question). Taper/race week: a morning daily-checkin through race day, the pacing card as a dashboard refresh, strava-sync after the dress rehearsal. Post-race: weekly-review of the whole build once the legs recover, plan-builder for the next goal, or nutrition-setup to reset targets for the off-season.
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
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.

