AlvaroLaraFF/strava-coach

/strava-coach

Skill to query Strava data and propose training plans. Detects user intent and executes the corresponding branch.

Pre-flight check

Before doing anything, verify the setup is complete:

python3 -c "
import sys; sys.path.insert(0, '.')
from strava.db import load_token
from strava.client import get_default_db_path
import os, time
errors = []
if not os.environ.get('STRAVA_CLIENT_ID'): errors.append('MISSING_CLIENT_ID')
if not os.environ.get('STRAVA_CLIENT_SECRET'): errors.append('MISSING_CLIENT_SECRET')
token = load_token(get_default_db_path())
if not token: errors.append('NO_TOKEN')
elif token['expires_at'] < time.time(): errors.append('TOKEN_EXPIRED')
print(','.join(errors) if errors else 'OK')
"

• If output is OK, proceed to intent detection.
• If there are errors, tell the user: "Setup is not complete. Run /strava-setup first to configure everything step by step." and stop.

---

Intent detection

Analyze what the user asked for:

• If they mention "activities", "latest", "recent", "what did I do", "workouts" → Branch A
• If they mention "statistics", "totals", "summary", "how much", "stats" → Branch B
• If they mention "propose", "train", "plan", "sessions", "next week", "what should I do" → Branch C
• If unclear, ask the user with AskUserQuestion.

---

Branch A — Recent activities

Step 1: Determine filters

Ask the user if they want to filter by sport or number of activities, or use defaults (last 20, all sports).

Step 2: Sync and query

python3 .claude/skills/strava-coach/scripts/list_activities.py --sync --limit 20

If the user requested a specific sport, add --sport-type Run (or Ride, Swim, etc.). If they requested a date range, add --days 30.

Step 3: Display results

Parse the JSON output. Show a formatted table with columns: Date, Type, Name, Distance, Duration, Pace/Speed, HR.

If the JSON contains "success": false, show the error to the user. If it's an authentication error, tell them to run /strava-setup.

---

Branch B — Global statistics

Step 1: Download statistics

python3 .claude/skills/strava-coach/scripts/get_stats.py --refresh

Step 2: Display summary

Parse the JSON and present:

• Year to date (YTD): distance by sport (running, cycling, swimming)
• All-time totals: cumulative distance by sport
• Last 30 days: total activities, distance, time, weekly frequency, sport breakdown

---

Branch C — Propose training sessions

Step 1: Sync recent activities

First make sure the database is up to date:

python3 .claude/skills/strava-coach/scripts/list_activities.py --sync --limit 50 --days 60

Step 2: Ask for goal

If the user didn't specify a goal, ask with AskUserQuestion:

• Prepare for a race (5K, 10K, half marathon, marathon)
• Improve speed/pace
• Maintain fitness / general health
• Lose weight
• Other (free text)

Step 3: Get analysis data

python3 .claude/skills/strava-coach/scripts/propose_sessions.py --days 30 --goal "GOAL" --next-days 7

Step 4: Generate proposal

Using the JSON data (weekly loads, pattern analysis, recent activities), reason as a sports coach and generate a weekly plan that includes:

• For each day of the upcoming week: session type, estimated duration, intensity (easy/moderate/hard/rest), exercise description
• Respect the user's current frequency and volume (don't increase more than 10% weekly)
• Include at least 2 rest or active recovery days
• If HR data is available, use training zones (Z1-Z5)
• Adapt to the stated goal

Present the plan as a table: Day | Type | Duration | Intensity | Description.

---

Error handling

| Error | Action | |-------|--------| | "success": false with auth error | Tell the user to run /strava-setup | | "success": false with API error | Show the Strava error message | | No activities in DB | Suggest using --sync to download |