log-session - SKILL.md Agent Skill

name: log-session description: Log a completed training session. Records what you did, compares to the program prescription, and updates your training log. Use after training.

Session Logger

Record a completed session to data/training_log.ndjson. Compare against prescription. Check for new PRs.

Inputs

Log date (today or explicit YYYY-MM-DD)
Session type/intensity and exercises performed (or as prescribed)
Readiness and context fields: RPE, sleep, soreness, optional notes
Program + PR reference files: data/program.json, data/prs.json

Outputs

One appended NDJSON log entry in data/training_log.ndjson
Updated PR values in data/prs.json when a new PR is detected
Logging confirmation output (including PR callouts if applicable)

Workflow

Step 0: Detect date and confirm log date

Read system date/time and show Today: YYYY-MM-DD.

Default log_date = today
If athlete says they are backfilling, use their provided date as log_date
If /checkin was missed, continue and log normally

Step 1: Ask the athlete (all at once)

Log date: today or YYYY-MM-DD (if blank/today, use detected date)
Session type: T / S / H / T2 / mini-T / mini-S / mini-H / cardio / recovery (+ intensity)
Exercises performed: what you did, or "as prescribed". Cardio: activity + duration + intensity.
RPE (1-10) | 5. Sleep: good/ok/bad | 6. Leg soreness (0-10) | 7. Notes (optional)

Step 2: Load program + compare

Read data/program.json. Calculate week number from log_date per CLAUDE.md formula (not always from today). Match performed exercises against prescription. "as prescribed" -> all completed. Unmatched extras -> bonus entries. Cardio -> no comparison.

Step 3: Build + save

Build entry as a single-line JSON object per schema in references/skill_schemas.md. Append it to data/training_log.ndjson using Bash: echo '<json>' >> data/training_log.ndjson. Append only - never read+rewrite the file.

Step 4: Check PRs

Read data/prs.json. Compare logged weights against PR-tracked exercises (CLAUDE.md "Personal Records"). If exceeded -> update prs.json.

Step 5: Output

Format per template in references/skill_schemas.md. Include new PRs if any.

Failure Modes and Fallbacks

/checkin not run earlier: continue and log normally
Unknown/unmatched exercise names: log as performed; mark unmatched as extras
Missing sets/reps in user input: use prescribed count when available
PR file missing/unreadable: still append session log; skip PR update and note it

Constraints

Append only - never read+rewrite the log file
If sets not specified, use prescribed count
Each entry must be valid JSON on a single line (NDJSON format)
If /checkin was missed, still log the session without warning/blocking