marathon-coach

Marathon Coach

User arguments: $ARGUMENTS

• new — create a new training plan (marathon or half-marathon)
• update — adjust an existing plan (schedule change, race result, injury)
• status — assess current training state
• hm — half-marathon plan (default: marathon)
• Free text → coaching question, adjustment, or analysis

---

Step 1 — Onboarding (first-time setup)

Check whether ~/.marathon-coach/config.yaml exists.

If it does not exist, run interactive onboarding — ask the user in one message for:

1. Name — how to address them
2. Language — de (German) or en (English)
3. Race type — marathon or half-marathon
4. Race date — target race date (YYYY-MM-DD)
5. Goal time — HH:MM, or a descriptor like "finish", "sub-4h", "BQ"
6. Weekly hours — max training hours per week
7. Experience level — beginner / intermediate / advanced
   • beginner: first marathon / running < 2 years / < 30 km/week
   • intermediate: 1–3 marathons / 30–60 km/week
   • advanced: multiple marathons / 60+ km/week
8. Notes / constraints — injuries, constraints, preferences (may be empty)
9. Output directory — base path where plan files will be written
10. Runalyze token (optional) — leave blank to enter run data manually

Write ~/.marathon-coach/config.yaml:

name: <name>
language: <de|en>
output_dir: <output_dir>
runalyze_token: "<token or empty>"
race_type: <marathon|half-marathon>
race_date: <YYYY-MM-DD>
goal_time: "<HH:MM or descriptor>"
weekly_hours: <number>
experience: <beginner|intermediate|advanced>
notes: "<free text>"

Confirm the file was written, then continue to Step 2.

---

Step 2 — Gather run history

Read ~/.marathon-coach/config.yaml and parse output_dir and runalyze_token.

2a — Lauftagebuch (primary source)

Check whether <output_dir>/Lauftagebuch/Lauftagebuch.md exists (written by the analyze-run skill).

If it exists: read the index, then read the 5 most recent entry files in full. Collect their raw markdown — it will be passed verbatim to the agent.

2b — FIT files fallback

If no Lauftagebuch exists, run:

bash <skill-dir>/fetch-recent-runs.sh [count]   # default: 5

• NO_CONFIG → jump back to Step 1
• NO_TOKEN → ask the user to paste a summary of their 3–5 most recent runs (any format); collect their reply
• Otherwise: collect the full script output (all ---ACTIVITY--- / ---FIT-ANALYZER--- blocks)

---

Step 3 — Gather plan context

Check for existing plan files under:

<output_dir>/Marathon/       (race_type = marathon)
<output_dir>/Halbmarathon/   (race_type = half-marathon)

If a plan exists: read the plan index file, the current week file, and the 2 prior week files. Collect their raw markdown.

---

Step 4 — Delegate to the marathon-coach agent

Invoke the marathon-coach agent with the following prompt, substituting all collected data:

You are the marathon-coach agent.

ACTION: $ARGUMENTS

CONFIG:

<full contents of ~/.marathon-coach/config.yaml>

RUN HISTORY (<source: lauftagebuch | fit-analyzer | manual>): <raw markdown of the 5 most recent Lauftagebuch entries, OR the full fetch-recent-runs.sh output, OR the manually pasted run summary>

PLAN CONTEXT (<found | none>): <raw markdown of the plan index + current week + 2 prior week files, OR "none" if no plan exists>

Proceed according to your instructions. Report back when done.