Controls and source defaults

Read this appendix correctly

This is a reviewed extraction of static control labels and AppStorage source initializers from fourteen explicitly selected Settings-related files. It is not a dump of preferences, accounts, device identifiers or secrets. Conditional and dynamically named controls are not all enumerated; they still require contextual review.

Literal initializers show the code fallback, not the current stored value. Expressions remain expressions so the reference does not guess a default that is resolved elsewhere. An empty string can represent unset/automatic configuration, not disabled functionality. Labels for buttons document available actions, not actions already performed. See Settings for the category map.

Example: a fallback is not your current value

commandem prefs default_mode

commandem prefs default_mode plan
commandem prefs default_mode --clear
commandem prefs default_mode

General settings controls

These controls span the twelve Settings categories above, including appearance/layout, chat defaults and system actions. Use Settings search for the exact label. Conditional controls may appear only after another option is enabled; a button label does not mean the operation already happened.

Reset appearance & behavior also restores the Home feed project, significance and maintenance filters, and the project-news significance and maintenance filters, to their defaults. Reset also restores saved resizable-section layouts. Pinned/hidden projects, shortcuts and Launch at Login remain unchanged.

Automatic continuation

This pane sets bounded turns, budget, model/effort, verification guard and progress reporting. Source initializers are 25 turns, 10.0 budget, default model/effort and verification guard true; actual stored settings can differ. Status progress is deterministic; optional model summaries can spend quota.

Quota exhaustion rules

Rules decide whether to wait, stop or use an explicitly permitted fallback chain. Check agent/account/model identity and the recorded exhaustion reason. A form placeholder or suggested wait duration is not an enforced source default.

Microphone and voice commands

Select an input device and dictation behavior. Device identifiers and OS authorization are local. Voice-command confirmation is a separate safety boundary; the UI does not grant unrestricted spoken execution.

Speech playback

Configure voice, rate, volume and auto-read where supported. Voice availability depends on the chosen playback implementation and installed resources; no universal cloud-free claim follows from this settings pane.

Backup and workspace export

Full backups and scoped workspace exports have different contents. Workspace exports exclude chats; a full state backup can include sensitive configuration. Inspect the archive and import scope, preserve backups and never publish private state by default.

Device pairing

Pair the intended companion device and inspect its trust record. Network/hardware availability is external to the source defaults; physical-device verification remains separate.

Notification policy

Per-type policy determines priority and delivery independently from durable unread state. Quiet/silent behavior is not deletion, and clearing a feed does not resolve the source project's health condition.

Reminders and Sync conflicts are separate per-type controls and source-default to Alert. Stored choices can override these defaults; OS permissions still gate banners and sound.

Provider account preferences

Keep provider defaults, named native accounts and endpoint profiles distinct. A serialized preference initializer is not authentication data. Inspect the selected account in the actual run receipt.

Roles and Loadouts

Role configuration supplies reusable account/tool/behavior references for supported dispatch paths. A Loadout binds names rather than copying credentials; validate bindings before unattended work.

Local role bindings use the shared discovered model catalog. Local agent/model choices carry the Local badge and available tier or estimate information; catalog metadata is not proof of live model readiness or safe tool use.

Protected capabilities

Configure the supported local capability broker and authentication factors. Passphrase-only operation has limits against same-UID access; tool approval and OS-factor authorization must remain explicit.

Policy editor

Inspect and edit the actual local execution policy. Saving a policy changes future permission decisions; a policy label is not independent proof of sandbox enforcement.

Standing directives

View/edit shared project instructions and apply them through the supported propagation mechanism. Applying can write project instruction files; review the diff and do not let an untrusted advisory change owner authority.

Dependency diagnostics

Inspect detected tools and their configuration. Installation, version discovery, runtime compatibility and authenticated access are different checks; a dependency row is not a provider certification.

No static label or stored initializer is declared in this selected source. Dynamic controls and child views are outside this extraction.