The shield titled itself with the first enabled rule whose selection
contained the app, regardless of whether that rule was the one actually
blocking. When several rules covered the same app (e.g. a dormant
schedule + the active limit) it named the wrong rule.
Drop the rule-name lookup entirely: every shield now carries a generic
"App Blocked" title. Open-limit shields keep their functional detail
(running count + "Open (N left)" button) beneath that title. The
text-only decision moves into a pure, unit-tested ShieldPresentation in
Shared/, and the now-dead ShieldLookup.snapshot(containingApplication:)
is removed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
RuleEnforcer.refresh only shielded a limit rule once its budget was
spent, so it tore down the background's proactive open-limit "turnstile"
shield (the shield is how opens are counted) whenever the app was
foregrounded, and a freshly created open-limit rule did nothing until
the next midnight reset.
refresh now applies the same proactive gate as
LimitEnforcement.handleDayStart: an enabled, scheduled-today, un-paused
open-limit rule is shielded even before its budget is spent, so it takes
effect immediately. A new app-group OpenSessionStore records a granted
"Open" session's expiry so neither enforcement path re-shields (and cuts
short) a sanctioned ~15-minute session.
Overlapping rules already enforce strictly: each rule shields its own
ManagedSettingsStore and Screen Time unions them, so an app is blocked
if any covering rule blocks it. Document that model in the spec (§4.8)
and lock it in with tests, including a time limit blocking during an
open-limit's granted session and a daily opens reset.
Co-Authored-By: Claude <noreply@anthropic.com>
Replace the wide BlockingRule/RuleDraft "god struct" — where every option
existed for every kind — with a RuleConfiguration sum type that carries only
the options each kind actually has:
.schedule(ScheduleConfig: window + selectionMode + blockAdultContent)
.timeLimit(TimeLimitConfig: dailyLimitMinutes)
.openLimit(OpenLimitConfig: maxOpens)
Name, days, Hard Mode, app list, and pause stay common to all kinds. This
makes illegal states unrepresentable: Block / Allow Only and Block Adult
Content are now structurally Schedule-only.
User-visible behavior change: the Time Limit and Open Limit editors no longer
offer a Block Adult Content toggle, and their detail sheets drop the "Adult
websites" row — those never made sense for a usage budget. Limit rules are
always Block and never engage the web-content filter.
BlockingRule keeps flat columns as raw persistence behind a computed
`configuration` bridge (lowest SwiftData risk; the cross-process RuleSnapshot
wire format and the Screen Time extensions are untouched). Logic, the editors,
and the detail sheet all switch on the sum type.
Spec (§1, §3.5/§3.6, §5.2) updated first; tests reworked to the new API with
new structural-guarantee unit tests and a UI test asserting the Time Limit
editor/detail omit adult content. Full suite green (180 tests).
Co-Authored-By: Claude <noreply@anthropic.com>
Time-limit and open-limit rules reuse BlockingRule's startMinutes/
endMinutes, which default to 09:00-17:00. Those fields are meaningless
for limit rules (the budget applies all day, resetting at midnight), but
status() still falls through to schedule.nextStart(), so an idle limit
rule reported .upcoming(startsAt: next 09:00) — rendered in the detail
sheet header as "Starts in 22h".
The home card already special-cased this; the detail sheet rendered the
raw status label, so the misleading countdown leaked there. Centralize a
kind-aware BlockingRule.statusLabel(for:relativeTo:) used by both the
card and the detail header: limit rules that aren't blocking show their
daily budget ("45m / day", "5 opens / day"); schedule rules and any
blocking/paused/dormant rule keep the live status label.
Verified on simulator: the Time Keeper detail header now reads
"Time Limit, 45m / day". 173 tests pass.
Co-Authored-By: Claude <noreply@anthropic.com>
Schedule (time-window) rules had no background enforcement: RuleScheduler.sync()
skipped them, so their shields were applied only by RuleEnforcer.refresh() — the
launch + 30s foreground loop. A window that began while the app was closed didn't
engage until the user reopened the app, which is why scheduled blocks could land
late or unevenly.
RuleScheduler now registers a repeating DeviceActivitySchedule per enabled
schedule rule's window (sched-<uuid>, plus sched2-<uuid> for windows that cross
midnight, since DeviceActivity can't express an interval whose end precedes its
start). The monitor extension routes these to a new ScheduleEnforcement.reconcile(),
which recomputes the rule's live schedule state from its snapshot
(RuleSchedule.isActive, honouring days, pause and the midnight-crossing rule) and
applies or clears the shield to match — the same logic RuleEnforcer.refresh runs
in the foreground, kept as the reconciliation safety net because interval callbacks
are known to fire late or not at all.
- RuleSnapshot gains startMinutes/endMinutes, with a tolerant decoder so snapshots
written before these fields still load instead of failing the whole batch and
blinding the extensions until the app reopens.
- Window activities are fingerprinted on their interval alone, so changing days,
mode or apps no longer needlessly restarts monitoring.
On-device verification of the background transition is still pending (the simulator
does not deliver DeviceActivity callbacks); covered by new unit tests across the
scheduler, the snapshot codec and the new enforcement reactions.
Co-Authored-By: Claude <noreply@anthropic.com>
- Shared/ layer compiled into the app and three new extension targets:
rule snapshots in the app group, the usage ledger, monitoring-plan
naming, and LimitEnforcement (shared, unit-tested event reactions)
- RuleScheduler mirrors rules to the app group and reconciles
DeviceActivity monitoring: one daily 00:00-23:59 activity per limit
rule, with a cumulative usage-threshold event per budget minute for
time limits; activities restart only when their configuration changes
(a restart resets threshold accounting)
- OpenAppLockMonitor (DeviceActivityMonitor): midnight budget resets,
records usage minutes, shields at the budget, re-shields when a
granted open session ends
- OpenAppLockShieldConfig: open-limit shields show 'Opened X of N times
today' with an 'Open (Y left)' secondary button
- OpenAppLockShieldAction: an Open press spends one open, lifts the
rule's shield, and starts the ~15-minute one-shot session
- extensions are classic NSExtension app extensions (the ExtensionKit
product type expects an @main entry and made the app fail to install);
shield-store tracking moved to app-group defaults so the app and
extensions see one consistent set
- maps iOS 26's new .approvedWithDataAccess authorization status to
approved (it previously fell through @unknown default to notDetermined)
- shares the OpenAppLock scheme (Xcode dropped the autocreated one when
targets were added)
Co-Authored-By: Claude <noreply@anthropic.com>
- RuleUsage + UsageLedger: per-rule, per-day minutes/opens in app-group
defaults (monotonic minutes, incrementing opens, midnight rollover by
day-keying); MockUsageLedger for tests and seeded UI scenarios
- usage-aware status: a limit rule whose daily budget is spent on an
enabled day is active (blocking) until next midnight; Hard Mode gating
and app-list locking honor it; soft unblock pauses until midnight
- RuleEnforcer shields spent limit rules (always Block mode) while the
app runs
- new Usage section under Blocked Apps: '18m of 45m used today · 27m
left', '2 of 5 opens today · 3 opens left', 'Blocked until tomorrow'
- new 'limits' seed scenario + UI tests
Co-Authored-By: Claude <noreply@anthropic.com>
- AppList @Model with launch migration from legacy inline selections
(rules with identical selections share one list; idempotent)
- rules point at one app list; Block/Allow Only belongs to the rule and
is offered only in the Schedule editor (limit kinds sanitize to Block)
- app-list picker sheet (select / create / edit / delete with in-use
protection) replaces the per-rule selection sheet
- Hard Mode locks app-list editing while any hard rule is blocking
- SwiftData stability: relationships are only wired between managed
models, and unit tests share one in-memory container (fresh context +
data wipe per test) — per-test container creation trapped
intermittently inside SwiftData
Co-Authored-By: Claude <noreply@anthropic.com>
Full rename ahead of App Store submission: project, targets, bundle
identifiers (dev.bchen.Severed -> dev.bchen.OpenAppLock), entitlements
file, app entry point, user-facing onboarding strings, test helpers,
and docs. Verified: all 62 unit tests pass.
Co-Authored-By: Claude <noreply@anthropic.com>
