Moved to https://github.com/kuhyx/build-your-x with full git history, where it
gained a `byox` progress tracker (crdt_sync-backed) and the builds themselves.
- git rm -r python_pkg/byox_ladder/
- CLAUDE.md: add build-your-x to the extracted-repos list
- .pre-commit-config.yaml: drop the now-stale byox_ladder codespell exclude
No remaining package imports byox_ladder.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Nw9Vw8ufxLjkDox7iXPMH3
Ranks the codecrafters build-your-own-x projects by difficulty and emits two
offline, self-contained HTML pages: a per-guide ladder (~360 tutorials) and a
per-category ladder (30 categories), grouped into Beginner/Intermediate/Advanced/
Expert tiers plus an Unsorted bucket. A guide inherits its category's tier;
language only nudges order within a tier. Includes a Makefile pipeline
(fetch -> parse -> build) and 100%-branch-covered tests. Excludes the generated
guide page from codespell since it inlines upstream titles verbatim.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QNRTPNYvoBTQoTEWkYtCJD
a12e696 turned CI red two ways, both missed locally because the suite was
run for the brother_printer package alone while CI runs everything in a
random order.
The repo caps Python files at 500 lines and cups_service.py had reached
659. Rather than cut it arbitrarily, extract consumable life tracking -
page counters, replacement baselines, the scale migration, dropped-page
detection - into consumables.py. The seam is real: none of it depends on
CUPS service control, and cups_service needs only estimate_consumable_life
for its fallback query, so the dependency runs one way. The test files are
split along the same seams.
test_main_entry passed or failed depending on execution order: runpy warns
when the target module is already in sys.modules, a sibling test imports
it, and pytest is configured with filterwarnings=error. Drop the module
before running it so the test no longer depends on who ran first.
Structural only; no behaviour change. 692 tests, 100% branch coverage.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RG3eEn8rwZbcXaJrKJxdv8
./run.sh failed outright with ModuleNotFoundError, and once running it
reported a sleeping printer as a critical paper jam. Both are fixed, plus
the defects found while verifying them against the real HL-1110.
Entry point: run.sh ran the script by path, so the absolute
python_pkg.* imports could not resolve. Resolve the repo root and pass
PYTHONPATH as a sudo argument, since sudo's env_reset strips an exported
one. __main__.py had the wrong import prefix and ran main() on import.
Status codes: the table claimed provenance from a "Brother PJL Technical
Reference" that does not appear to exist, and 23 of its 25 entries were
wrong. Rebuilt from the HP PJL Technical Reference Manual, Appendix D,
and verified against six states induced on real hardware (10001 READY,
10003 WARMING UP, 10023 PRINTING, 40000 SLEEP, 40021 TOP COVER OPEN,
41213 NO PAPER) - all six matched the reference, none matched the table.
40000 is SLEEP MODE, which the reference explicitly flags "not an error";
the real jam code 40022 was absent entirely. The inverse also bit: 10006
is TONER LOW, previously shown as a benign "Processing". Codes in no
reference are dropped and now report "Unknown status - check printer
display" rather than a confident wrong answer, and states we infer
ourselves no longer borrow PJL numbers to describe themselves.
Page counts: estimates came from the CUPS page log, which only sees jobs
CUPS logged and counts pages the printer never printed (it logged
total=50 for a job that printed 3). Take them from @PJL INFO PAGECOUNT
instead, migrating stored baselines onto the printer's scale once so the
reported percentages do not jump.
Reading status must not cost anything: a blocking open() hung forever
whenever the printer was unwell - precisely when a status report is
wanted - and the port-status probe stopped CUPS and reset the USB device,
destroying the in-flight job and then misreporting the damage as a
printer error. The fd is now non-blocking with deadline-bounded writes,
and the probe reads the same byte via the free usblp ioctl, skips
entirely while a job prints, and re-attaches the kernel driver.
Dropped pages: the HL-1110 silently discards pages whose 600dpi raster
exceeds its ~1MB memory, stays READY, and CUPS still calls the job
successful - so nothing anywhere reports the failure. Compare CUPS's
claim against the printer's counter between jobs and name the workaround
(lp -o Resolution=300dpi).
402 tests, 100% branch coverage.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RG3eEn8rwZbcXaJrKJxdv8
test_codebase_fingerprint_stat_oserror_raises passed on 3.14 but failed
on CI's 3.11: there Path.is_file() itself calls stat() and re-raises an
OSError without a recognized errno, so the injected error escaped before
the loop reached the guarded path.stat() call. Force is_file() True for
the target (without calling stat) so the explicit fingerprint-time stat()
deterministically raises and hits the except OSError branch on every
Python version. Verified passing on 3.11.11 and 3.14.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RRCdR4D8SaVrN2xraMQxM
New python_pkg/code_tutor package that turns an unfamiliar codebase into
an interactive lesson: analyse code into items (_analyzer), order them
dependency-first (_plan_builder), pose reimplement challenges
(_challenge + _challenge_support), verify user implementations with tests
and coverage (_verifier), track progress (_progress/_session), and offer
local-LLM hints (_llm), driven by a Typer/Rich CLI (cli).
- meta/pyproject.toml + .pre-commit-config.yaml + meta/requirements.txt:
add anthropic/rich/typer deps, a cli.py per-file-ignore (Typer B008 etc),
and code_tutor-specific ruff/pylint/mypy tuning.
- _challenge.py and its tests are split across _challenge_support.py and
test_challenge_part2/part3.py to stay under the 500-line file cap.
Verification: pytest 631 passed at 100% line+branch coverage; every file
<=500 lines; pre-commit (ruff ALL, mypy strict, pylint 10/10, bandit,
evidence + multi-file contract) green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RRCdR4D8SaVrN2xraMQxM
Full history preserved via git filter-repo; production
wake-alarm.service already cut over to the pip-installed standalone
package and verified clean.
Also fixes morning_routine._orchestrator, which has been silently
broken since screen-locker's extraction on 2026-05-28: it still
referenced python_pkg.screen_locker.screen_lock, causing
ModuleNotFoundError on every morning the workout-lock handoff ran.
Both module references now point at the pip-installed standalone
packages (wake_alarm._alarm, screen_locker.screen_lock); fixing this
required pip-installing screen_locker into system Python for the
first time too (see the separate screen-locker packaging-fix
commits), which it had never been.
Full history preserved via git filter-repo; production
diet-guard-gate.service already cut over to the pip-installed
standalone package and verified clean. No other monorepo file
referenced diet_guard, so removal is a clean delete.
The previous two fix attempts (removing -n 4 xdist parallelism, ruling
out a Python 3.12-vs-3.14 difference) both targeted the wrong cause for
the coverage gap on _alarm_display.py:71. The actual bug:
_restore_display() was never mocked in the _block_extra_devices fixture
shared by test_alarm_part2/3/4.py, so every test exercising
on_close()/_lock.close() ran the REAL _restore_display(), which calls
the REAL shutil.which('xset'). My dev machine has /usr/bin/xset
installed, so it always hit the "found" branch by accident; the CI
runner doesn't, so it always hit the "missing" branch instead, and
there was no dedicated unit test for the "found" path at all.
Fix: mock _restore_display in that fixture (matching its already-mocked
siblings _restore_fans/_restore_alarm_audio/turn_off_plug), and add a
dedicated, hermetic test for the xset-found branch that doesn't depend
on whether the running machine actually has xset on PATH. Side benefit:
the test suite no longer makes a real `xset s on` subprocess call on
the developer's desktop as an untracked side effect of running tests.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01A7vbgtFfZmfxJtN5DdtJky
test_demo_opens_window and test_bare_gate_due_opens_window never mocked
wait_for_display(), so they silently relied on a real, reachable X
display being present. On a dev machine this is always true, so they
passed; on a true headless CI runner (no X server at all) each test
blocked for the real 60s timeout before failing, explaining the CI
run's 131s duration. Mocking it to True (matching the existing
test_gate_due_but_display_not_ready_defers pattern for the False case)
restores hermeticity with no production code change.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01A7vbgtFfZmfxJtN5DdtJky
Third and final leg of the diet_guard -> screen-locker -> wake_alarm
gatelock extraction. WakeAlarm now composes gatelock.GateRoot +
LockWindow(mode="soft") instead of driving tk.Tk() directly, and moves
hardware teardown into on_close() so it runs on every exit path
(including SIGTERM), closing the prior gap where killing the process
left fans maxed. _state.py's HMAC import moves to gatelock.log_integrity,
the last call site of python_pkg.shared.log_integrity, which is deleted.
Manually verified on the real display: fullscreen overlay renders,
xdotool-typed input reaches the dismiss-code Entry (mode="soft" keeps
the existing typing-focus behavior), and SIGTERM exits within ~0.4s
while restoring hardware state. Full repo suite: 949 passed, 100%
branch coverage.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01A7vbgtFfZmfxJtN5DdtJky
MealGate now composes gatelock.GateRoot + gatelock.LockWindow instead of
inheriting the deleted _GateWindow/_GateRoot, and its HMAC signing goes
through gatelock.log_integrity. This is the first of three migrations
(diet_guard -> screen-locker -> wake_alarm) extracting the lock-window
mechanics that diet_guard's own _GateWindow proved out into a shared,
reusable package. Window-mechanics tests moved with the code; diet_guard's
suite now only tests its own wiring (LockConfig choice, hook delegation).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XCdT46zV8hESDvbgYMGDLt
Split diet_guard/_gatelock.py, wake_alarm/_alarm.py, and the
usage_report.py/_usage_report_parsing.py pair into focused
sub-modules so every Python file is <= 500 lines, satisfying
test_file_length.py. Install python-kasa into .venv (declared in
requirements but missing after the 3.13->3.14 venv upgrade),
fixing 8 failing smart_plug tests and restoring 100% coverage.
Also includes prior in-progress work from the working tree: the
wake_alarm Progress/View/Hardware field-grouping refactor,
brother_printer query module + tests, diet_guard foodbank/state/cli
updates, new shared coerce/logging_setup helpers, morning_routine
orchestrator tweaks, dwm window-manager config, gaming scripts, and
misc maintenance/digital-wellbeing script updates.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Add the diet_guard package: a screen-locking meal-logging gate that fires
on 4-hour slots (08/12/16/20) and records calories/macros, persisting an
autocompleting food bank.
- Trigger fix: the systemd timer fires at session start (Persistent=true)
before lightdm has written ~/.Xauthority, so the gate crashed with a
TclError instead of locking the screen. Add wait_for_display() /
_display_is_ready() in _gatelock.py and wire it into _cli._cmd_gate so the
gate retries on the next tick instead of crashing; add
Environment=XAUTHORITY=%h/.Xauthority to the service as belt-and-suspenders.
- Food-bank hardening: a transiently corrupt food_bank.json was warned about
on every keystroke and then silently overwritten (data loss). _read_bank
now quarantines it via _quarantine_corrupt_bank() (warn-once + timestamped
backup) before starting fresh.
- Multi-item meals: new _meal.py (MealItem, meal_total, MEAL_SOURCE),
remember_meal() + _upsert() in _foodbank.py, and a "+ Add item" control in
the gate that logs both the individual items and the composite meal.
- Bundle resolve_nutrition's manual macros into a ManualMacros dataclass to
stay within the argument-count limit.
diet_guard at 100% branch coverage; full pre-commit suite passes.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Split _alarm.py (1059 lines) into _alarm.py + _audio.py + _challenges.py
- Split test files (1305 / 725 lines) into 6 files, all under 500 lines
- Replace random.* with secrets.* (S311); fix RUF001, SIM117, E501 ruff errors
- Rewrite pytest_changed_packages.py to always run all packages with global
--cov python_pkg coverage (100% branch coverage enforced across whole tree)
- Add DISMISS_ROUNDS_REQUIRED=2 and DISMISS_FLASH_SECONDS=4 to _constants.py
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Moved to https://github.com/kuhyx/screen-locker with full git history,
vendored dependencies (shared.log_integrity, wake_alarm constants +
has_workout_skip_today), 392 tests at 100% branch coverage.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Moved to https://github.com/kuhyx/steam-backlog-enforcer with full git
history, rewritten imports, standalone pyproject.toml, and CI.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Four bugs fixed:
- HLTB search returned 0 results for ~87 games with special chars (™, ®, &,
standalone -, (Legacy), RHCP, etc.) — add _sanitize_search_name() and
extend _build_search_variants() with Steam-suffix and edition stripping
- fetch_hltb_detail_missing returned immediately because `app_id not in rush`
was always False (all keys present with -1) — fix to `rush.get(id,-1) <= 0`
- save_hltb_cache overwrote rush/leisure on confidence-only partial saves —
now reads existing cache and preserves data when extras dicts are empty
- _filter_qualifying_games excluded 57 games with stale snapshot hours (-1)
even though HLTB hours cache had valid data — add cache fallback
Result: stats shows Rush 64,670h / Leisure 136,807h / Worst 228,594h for
all 785 qualifying games with full rush+leisure detail.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- New python_pkg/morning_routine package: sequential orchestrator runs wake
alarm then workout lock as blocking subprocesses (one fullscreen owner at
a time). Deployed as morning-routine.service; sleep hook updated to start
it on hibernate-resume instead of the standalone wake-alarm.service.
- wake_alarm: force G27Q HDMI card profile on at alarm time, poll up to 6s
for sink to appear, set as default + unmute 100%. Alarm now persists until
the typeable code is entered (no more silent 30-min give-up). Service gets
DISPLAY=:0 + ExecStartPre sleep 1 to fix cold-boot Tkinter crash.
- phone_focus_mode/config.sh: whitelist Revolut, mObywatel, VaultKitBypass.
- 100% branch coverage maintained across wake_alarm and morning_routine.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Drop overrideredirect on the dismiss UI: on X11 it bypassed the WM and
silently killed keyboard focus on the Entry, so the user could not type
the dismiss code. -fullscreen + -topmost still cover the whole screen.
- Add _max_sink_volume() / _restore_sink_volume(): unmute the default
PulseAudio/PipeWire sink and raise it to 100% at alarm start, restore
the original volume + mute state on dismiss. This is the biggest audio
lever — pcspkr is hardware-fixed and was already maxed.
- pcspkr: 3 back-to-back 1.5s beeps per cycle (was 1x 0.8s) at near-S16
max amplitude.
- Fans script: control every NCT pwm[1-9] channel, persist per-channel
pre-alarm state to /run/wake-alarm-fans.state so restore needs no args.
- Tests: 100% branch coverage on python_pkg/wake_alarm/ (140 passed).
Replaces the auto-reassign-to-shorter-game logic (which fired while the
current game was still in progress) with a strict workflow:
1. Check if assigned game is finished.
2. If not, do nothing.
3. If yes, pick the next shortest game and prompt the user.
4. If the user skips, ignore that game for 7 days and pick the next
shortest candidate.
Changes:
- State: add skipped_until + skip_for_days + active_skipped_ids.
- scanning.pick_next_game: optional on_select callback drives a
sequential picker that filters skipped IDs; legacy cmd_pick flow
preserved when on_select is None.
- _cmd_done._finalize_completion: pick + prompt via on_select.
- _cmd_done: remove _try_reassign_shorter_game and helpers
(_apply_cached_confidence_to_games, _should_reassign_candidate,
_echo_reassign_decision, _evaluate_reassign_iteration) plus call
site in cmd_done.
- Tests: drop obsolete _try_reassign_shorter_game suite; add
TestPromptKeepOrSkip, TestPickNextGameSequential, and State
skipped_until tests.
- Add python-kasa-based smart-plug control (_smart_plug.py) with
turn_on_plug / turn_off_plug called around the alarm window.
Reads ~/.config/wake_alarm/tapo.json (host/email/password).
- Hard timeout (TAPO_TIMEOUT_SECONDS) so plug never blocks the alarm.
- Install fan-control script + sudoers entry (install.sh step 6);
_max_fans / _restore_fans now invoke it via /usr/bin/sudo -n so
pwm1_enable writes succeed.
- Remove ntfy.sh push notifications entirely (silent no-op was useless).
- Replace every silent skip with _logger.warning() so failures are
loud: missing xset / xrandr / speaker-test, unreadable hwmon files,
fan script errors, missing Tapo config, kasa import failure, etc.
- wake-alarm.service: Restart=on-failure with 10s backoff.
- Tests: 100% line+branch coverage on python_pkg/wake_alarm.
- install.sh: install sleep-hook.sh to systemd-sleep hooks (step 3) and
shutdown-wrapper.sh to /usr/local/bin/shutdown (step 5)
- shutdown-wrapper.sh: new script that intercepts shutdown calls and
redirects to rtcwake -m disk on alarm nights (Mon/Fri/Sat/Sun), pass-
through for reboots and cancel commands
- sleep-hook.sh: new systemd-sleep hook that restarts wake-alarm.service
for each logged-in user after hibernate resume
- setup_midnight_shutdown.sh: check wake-alarm day; if yes set RTC alarm
and hibernate, otherwise fall back to systemctl poweroff
- Added SCHEDULED_SKIPS_FILE constant pointing to scheduled_skips.json
- Added _is_scheduled_skip_today() method: reads JSON list of YYYY-MM-DD
strings, exits 0 if today's UTC date is found (skips lock entirely)
- _shutdown.py: changed rtcwake -m no -> -m disk so machine hibernates
immediately when scheduling morning alarm (bedroom use)
- Added tests/test_scheduled_skip.py with full branch coverage
- Added scheduled_skips.json with initial skip dates
- library_hider.py: add safeHide(ids) JS helper that binary-bisects on failure
to skip problematic DLC/tool IDs without blocking the entire hide pass
- library_hider.py: increase CDP timeout 30s -> 120s; extract richer CDP error
details from exceptionDetails/exception.description
- _hltb_detail.py: rewrite _extract_base_leisure_hours() to pick the maximum
(slowest) time across all platform comp_high values and *_h fields; add
_platform_comp_high_candidates() helper
- Remove skip_app_ids from user-editable Config; callers updated
- Split PROTECTED_APP_IDS: only Steam infra/Proton IDs remain; game
IDs moved to a new time-locked exception system
- Add _whitelist.py: 24-hour cooldown on new exceptions, entropy-
checked justification (>= 5 words), append-only audit log,
chattr +i immutability on enforcement-critical config files
- Add is_protected_app() in game_install.py; used everywhere
instead of direct PROTECTED_APP_IDS membership checks
- Add 'add-exception' CLI command (cmd_add_exception in main.py)
- Call promote_pending_exceptions() and lock_enforcement_files()
in each _enforce_loop_iteration
- 590 tests, 100% branch coverage on all steam_backlog_enforcer modules
- Add .worktrees to .gitignore
- linux_configuration/tests: update script paths after periodic_background/
reorganisation (hosts_file_monitor, makepkg_capped, music_parallelism,
shutdown_timer_monitor, usage_monitoring_installer_efficiency)
- test_i3blocks_efficiency.sh: remove checks for HEARTBEAT_INTERVAL_S and
WARP_POLL_INTERVAL_S constants that no longer exist
- test_pacman_wrapper_security.sh: remove tests 20-21 (builtin time helpers /
external date calls) that are no longer applicable; update path
- generate_hosts_file.sh: add sed unblock rules for delio.com.pl and
loverslab.com to stay consistent with install.sh whitelist
- steam_backlog_enforcer/scanning.py: remove unplayable_reason arg from
logger.info call (too many format args); drop matching test assertion
- steam_backlog_enforcer/tests/test_protondb.py: add
test_unplayable_reason_no_trending_tier to restore 100% branch coverage
on protondb.py line 97 (was previously covered indirectly)
- Screen locker: disable VT switching (Ctrl+Alt+Fn) via setxkbmap
srvrkeys:none on startup; restore on close (production mode only).
Gracefully skips if setxkbmap is not installed (shutil.which).
Tests: 7 new tests, 100% branch coverage maintained.
- Midnight shutdown: restore real schedule values (Mon-Wed 21:00,
Thu-Sun 22:00, morning end 05:00); re-enable the three commented-out
leniency checks in check_schedule_protection(); self-lock script with
chattr +i at end of enable_midnight_shutdown().
- Hosts install: add UNBLOCK_STATE_FILE tracking for whitelisted domains;
check_unblock_entries_protection() blocks installation if the unblock
list grows; save state after install; self-lock install.sh and
generate_hosts_file.sh with chattr +i.
- meta/.pre-commit-config.yaml: move pytest-coverage hook to pre-commit stage
- scripts/pytest_changed_packages.py: single batched pytest -n auto invocation
with one --cov flag per affected python_pkg subpackage, wrapped in
systemd-run --user --scope -p MemoryMax=4G -p MemorySwapMax=0 when available
- python_pkg/steam_backlog_enforcer/tests/conftest.py: new autouse
_no_real_sleep fixture patches time.sleep across game_install /
library_hider / steam_api / _enforce_loop. Removes 3x 15s real sleeps
in TestFinalizeCompletion that fired through _ensure_steam_running
steam_backlog_enforcer test wall time: 33.97s -> 5.61s (xdist, no-cov)
5-package batched run: 732 tests in 1.37s @ 668% CPU
Coverage stays at 100% on all affected packages.
Evidence: docs/superpowers/evidence/pre-commit-pytest-batch-2026-05-14.json
- Move pyproject.toml, .pre-commit-config.yaml, requirements.txt, run.sh,
lint_python.sh, .fvmrc into meta/ with root symlinks preserving tool
auto-discovery.
- Combine requirements.txt + requirements-dev.txt into meta/requirements.txt
(single sorted source of truth).
- Remove setup.sh, .binary-allowlist, C/ (no native code remains),
python_pkg/{split,pdfCentered,geo_data}, scripts/check_c_cpp_build_files.sh.
- Drop clang-format/cppcheck/flawfinder/check-c-cpp-build-files hooks and
archived path excludes from pre-commit config.
- Add .secret-patterns to .gitignore and untrack it (sensitive content;
full history purge is a follow-up step).
Adds 1410710, 10500, 813780, 489830 to PROTECTED_APP_IDS so the enforcer
will not uninstall them. Existing tests patch the set, so test outcomes
are unaffected.
Adds a sick-day exemption flow with debt tracking so workout enforcement
can be skipped on declared sick days while preserving phone-verification
and shutdown invariants.
- New _sick_tracker module persists sick_history.json (days, debt, commitments).
- New _sick_dialog integrates declaration into the lock UI flow.
- _ui_flows.py and screen_lock.py consult tracker before enforcing workouts.
- gitignore sick_history.json (runtime state, like sick_day_state.json).
- 304 tests pass; 100% branch coverage on every screen_locker file.
- Set oom_score_adj=1000 in git hooks so OOM killer targets
pre-commit first, never crashing the PC
- Cap Node.js heap to 512MB for eslint/prettier/vitest
- Remove broken systemd-run cgroup wrapper (didn't work)
- cppcheck: process files one-at-a-time, --check-level=normal
- pytest: run packages sequentially in separate subprocesses
- Remove --force from cppcheck (exponential memory on #ifdef combos)
systemd-run --scope -p MemoryMax=3G per pytest subprocess so each
package gets its own memory cap, freed completely before the next.
Also use shutil.which + pathlib per ruff rules.
