← ClaudeAtlas

weekly-adaptlisted

Weekly training-plan adaptation loop based on logged actuals
seungwee-choi/oh-my-personal-best · ★ 3 · AI & Automation · score 74
Install: claude install-skill seungwee-choi/oh-my-personal-best
<Purpose> weekly-adapt is the weekly adaptation loop. It runs every week to compare what was planned against what was actually done, assess fatigue and compliance, and adapt next week's plan accordingly — advancing, holding, or deloading based on real data. Because it runs on a weekly cadence, the plan stays alive and responsive rather than static. A plan that ignores what happened last week is not a plan; it is a wish list. </Purpose> <Use_When> - The runner checks in at the end of or during a training week ("이번 주 계획 조정해줘", "weekly check-in", "adjust this week", "how did I do?") - Keyword triggers detected: "weekly", "이번 주", "adjust", "week check-in", "this week" - A new week is starting and the runner wants next week calibrated to actual last-week performance - plan-state.json exists and `critic_approved: true` (an active plan is in progress) - After recovering from illness or a missed-session cluster and the plan needs resetting </Use_When> <Steps> ## Step 1 — Collect and Aggregate Actuals (data-logger) Invoke `oh-my-personal-best:data-logger` to collect and aggregate this week's training data. If the runner provided check-in notes in `$ARGUMENTS`, pass them to data-logger for ingestion first (natural-language path). data-logger normalizes any new entries into `training-log.jsonl`. data-logger then produces the weekly summary: - Planned sessions (from plan-state.json `key_sessions`) vs. actual sessions logged - Compliance rate: sessions completed / sessions planned