rlacombe/switchback-running/review

Review a Workout

Minimum evidence

Identify the requested activity. Retrieve its compact detail and the matching planned event; use both providers only when each adds material evidence. Read the profile, relevant activity memory, and the knowledge chapter for the workout's purpose. Retrieve streams only for a defined analysis such as pace drift, heart-rate decoupling, intervals, or climbing.

Use get_activities or get_strava_activities to identify the session, then get_activity or get_strava_activity for detail. Use get_event for the full planned prescription. Request only the stream types needed for the named analysis.

Analysis

Count duplicate provider records once. Distinguish planned and completed values. Report material differences in distance or time, elevation, intensity, and load; use 10% as a review trigger interpreted in context. Verify units and calculations. Separate provider facts, athlete narrative, derived metrics, and hypotheses.

Output and memory

Explain whether the workout served its purpose, what the evidence supports, what remains uncertain, and what should change next. Avoid unsupported praise or diagnosis. When activity memory is enabled, archive the workout using agents/activity-memory.md and the activity template. Preserve athlete-written text and label the companion assessment as a hypothesis or decision where appropriate.