screen-locker/README.md

screen-locker

Tkinter/systemd screen locker with workout tracking, shared sick days, and wake-alarm integration. It enforces a workout cadence by locking the screen until a verified (RunnerUp, StrongLifts) or budgeted manual workout is logged.

Sick days

A sick day is a date in the fleet-wide sickdays pool, nothing else: no budget, countdown, justification or debt. The lock screen's I'm sick button (retry screen, no-phone screen and the Tue-Thu prompt) marks today with one tap and unlocks; sickdays add or the phone do the same. On a sick day the workout lock and the grace floor stand down, wake-alarm's workday penalty is void, and shutdown is max(20:00, the normal ladder) capped at 23:00 (earned_time.resolve(..., sick=True)). A day marked after its daily reset is lifted to 20:00 by the next 5-minute pass (_sick_floor); a workout that day still earns normally and is paid out of that lift first, so it lands on the ladder value (e.g. 20:40), never 20:00 plus its whole reward. install.sh migrates the old sick_history.json dates into the pool once (scripts/migrate_sick_history.py).

A sick day may keep lockers (sickdays add --keep workout). A kept workout lock behaves exactly as on a normal day (lock, quota, relaxed-day penalty on that date); only the 20:00 shutdown floor still applies, since it belongs to the day, not to a locker. The one-tap button keeps nothing. The missed-alarm penalty is void on its ring day only if alarm is off, and on its penalty day only if workout is off. All of it goes through _sick.stands_down_workout.

Shutdown time

Every day starts at 19:00 from 2026-10-01, 20:00 before it (base_minutes() in screen_locker/_shutdown_base.py) and bonuses push it later, capped at 23:00: the first counted workout +2h, each further one +1h, an accepted LeetCode submission that day +1h (flat, read from leetcode-guard's ledger, once per day), and a credited paper-book reading session of 20+ pages / 20+ minutes +1h (flat, read from book-guard's ledger, once per day, keyed on when the reading ended). The config is re-derived from those sources on each new day, so nothing has to be "remembered" across reboots.

A RunnerUp run is credited from its TCX export, found in the WebDAV drop directory RunnerUp uploads to on this machine (~/data/cloud/RunnerUp/, plus processed/ once endurain-import has taken it) or, when a phone is adb-visible, in /sdcard/Documents/RunnerUp/. The 15-minute workout-sync.timer scans both, so a run finished after 20:00 pushes the hour later without the phone being plugged in; the per-minute shutdown check then lifts the night lockdown on its own once the hour is past "now" (digital-wellbeing's check script).

MCP server (Claude Code integration)

screen-locker exposes a read-only MCP server (screen_locker._mcp) so Claude Code and its subagents can query workout compliance and the lock-decision state through typed tools — without shelling out to the screen-locker-status CLI or opening the Tk window.

• Read-only tools: get_status (full status snapshot), get_summary (one-line i3blocks summary + ok/warn/lock state word), explain_lock (why the screen is / isn't locked right now), get_flags (the individual boolean lock-decision predicates for today).
• There are no write/action tools, by design. Nothing here logs a workout or mutates state — workouts are logged only from RunnerUp-verified TCX data, never from a caller's claim. No tool exposes the sync token or any HMAC key.

The mcp SDK is an optional dependency (pip install -e '.[mcp]'), kept out of the CLI/systemd system-python path. One-time setup of the dedicated venv that Claude Code spawns:

./scripts/setup_mcp.sh

Registration lives in the checked-in .mcp.json (project scope). Restart Claude Code in this repo and approve the project MCP-server prompt for it to load. Verify with claude mcp list.