time-calibrate

<!-- tested with: claude code v2.1.122 --> # /time-calibrate re-measure your personal throughput and produce a diff report against the generic baseline in `plugins/cc/rules/time.md`. ## what to do when invoked 1. **check for `~/.claude/lore.db`** via `Bash`: ```bash test -r "$HOME/.claude/lore.db" && echo present || echo absent ``` 2. **if absent**, explain and stop: - the calibration needs session history in `~/.claude/lore.db`. - that database is built by the `lore` plugin from `~/.claude/projects/`. - install via `/plugin install lore@anipotts`, run `/lore:graph`, then re-run `/time-calibrate`. - do not fail loudly. one-screen message, no stack traces. 3. **if present**, resolve active effort (same precedence as `/time-estimate`) and current model, then run the calibration SQL via `Bash` + `sqlite3 -readonly`. ## calibration SQL paste into `sqlite3 -readonly ~/.claude/lore.db <<SQL ... SQL` or use a heredoc. all queries are parameter-free and read-only. ```sql -- A. session-length bucket distribution SELECT CASE WHEN duration_wall_seconds < 300 THEN '1_under_5min' WHEN duration_wall_seconds < 900 THEN '2_5to15' WHEN duration_wall_seconds < 1800 THEN '3_15to30' WHEN duration_wall_seconds < 3600 THEN '4_30to60' WHEN duration_wall_seconds < 7200 THEN '5_60to120' ELSE '6_over120' END AS bucket, COUNT(*) AS n, ROUND(AVG(duration_wall_seconds)/60.0, 1) AS mean_wall_min, ROUND(AVG(duration_active_seconds)/60.0, 1) AS...