Analyze Ride

Pipeline

0 — Determine workout source

If the user explicitly mentions gym/strength/weight training, use the Strava gym pipeline below. Otherwise, check TrainingPeaks first. If the workout isn't in TrainingPeaks (e.g., gym sessions), fall back to Strava.

Strava gym pipeline: For gym/strength workouts, use Strava instead of TrainingPeaks:

1. get-activities to find the session (filter by date or pick most recent non-cycling activity)
2. get-activity-details for duration, HR data, activity name, description
3. get-activity-streams with keys ["heartrate", "time"] for HR time-series if available
4. Skip the timeseries analysis script (Step 3) — it's cycling-specific
5. File with gym-appropriate frontmatter: sport: Gym, no power metrics, include duration_minutes, average_hr, max_hr, rpe
6. Continue from Step 5 (Check prescription) onward

For cycling workouts, continue with Step 1 below.

1 — Identify the workout

If the user specifies a date or name, use that. Otherwise, pull the most recent ride via tp_get_workouts for the last 3 days. If multiple workouts exist, list them and ask which one.

2 — Pull workout data

Run in parallel:

• tp_analyze_workout for detailed analysis (time-series, zones, laps)
• tp_get_athlete_settings for current FTP and zones

tp_analyze_workout returns two things:

1. Inline data — totals (TSS, NP, IF, VI, EF, distance, elevation), dataElements (channel metadata with zone definitions), lapData (per-lap summaries)
2. Raw timeseries file — written to disk at a path like /var/folders/.../tp-mcp/analysis/workout_{id}.json. Contains data array with rows at ~4-second intervals, each with: time, Power, HeartRate, Cadence, Speed, Altitude, Grade, Distance, Torque, Structure (target % FTP). This file is typically 200-400KB and cannot be read directly into context.

Capture the data file path from the response — you'll need it for Step 3.

3 — Run timeseries analysis

Use the analysis script to process the raw timeseries file:

python3 <skill-path>/scripts/analyze_workout.py <data_file_path> <command>

Available commands:

Recommended approach:

• Start with all for a comprehensive view
• Or be targeted: intervals for structured workouts, fatigue for long rides, pacing for races/group rides

4 — Check for PRs

Use tp_get_workout_prs to see if personal records were set during this workout.

5 — Check prescription

Read the current training plan page(s) from the knowledge base (plans/tp-*.md) to find what was prescribed for this date. If no prescription exists, note that.

6 — Pull recovery context (optional)

If Apple Health MCP is connected, query for the same date:

• HRV on waking
• Sleep quality/duration the night before
• Resting heart rate

If data is unavailable or stale, skip and note the gap.

7 — Gather weekly context

Use tp_get_weekly_summary to see how this workout fits into the week's overall training load.

8 — Load historical context

• Read wiki/index.md to find the 3-5 most recent workout analyses (analyses/wa-*.md)
• Read those pages to compare patterns (same workout type, similar TSS, trends)
• Read wiki/overview.md for current fitness context

9 — Discuss takeaways

Present the analysis to the athlete using the structured output format below. Combine the inline MCP data, script output, PRs, prescription comparison, and historical patterns into a coaching narrative.

This is a conversation checkpoint — wait for the athlete to respond before filing.

10 — File the analysis

Create analyses/wa-<slug>.md in the knowledge base. Slug format: YYYY-MM-DD-<descriptive-name> (e.g., wa-2026-04-12-saturday-group-ride).

Include full frontmatter:

---