feat(settings): let an MDM profile set a minimum safe mode level

[datlechin]

The first piece of managed configuration: an administrator can set a minimum Safe Mode level through a macOS configuration profile, and the user cannot drop below it.

The mechanism is smaller than it sounds, on purpose

macOS already merges a configuration profile's values into UserDefaults at the highest priority, so a forced key simply wins with no special read path. The only thing the ordinary API cannot answer is whether a key was forced, and CFPreferences.h says exactly what that is for:

In cases where machines and/or users are under some kind of management, callers should use this function to determine whether or not to disable UI elements corresponding to those preference keys.

So this adds a flat key namespace and a reader, not a policy engine, an override chain, or a second settings store. ManagedPolicyReader.isManaged exists to disable a control, not to fetch a value.

Why a separate flat key rather than forcing an existing setting

The app's settings are stored as encoded structs under keys like com.TablePro.settings.mcp. Forcing one of those would mean an administrator had to reproduce the entire struct in their profile, and it would break the first time a field was added. Policy keys are one primitive per decision:

Key	Type
com.TablePro.policy.minimumSafeModeLevel	String
com.TablePro.policy.mcpServerDisabled	Bool
com.TablePro.policy.aiAssistantDisabled	Bool
com.TablePro.policy.pluginInstallDisabled	Bool

Only the first is enforced in this PR. The other three are declared because they are the same shape and the enum is the place a reviewer should see the intended set, but nothing reads them yet, so treat them as reserved rather than shipped.

Two behaviours worth reviewing

The policy is a floor, never a ceiling. A connection set stricter than the policy keeps its own level. Someone who put a production connection on Read-Only still gets Read-Only when the policy only asks for Alert. ManagedPolicyResolverTests checks every ordered pair of the six levels in both directions, so no pair can silently collapse to a tie.

An unrecognised value imposes no floor at all. Not Read-Only, which would lock people out over a typo in a profile, and not Silent, which would drop the policy without saying so.

It is enforced at ExecutionGateProvider's safeModeLevelResolver, which is the single place the gate learns a connection's level, so every caller including the AI assistant and the MCP tools goes through it.

Scope

Safe Mode is a floor on TablePro's behaviour, not on the database. It stops the app issuing a write; it does not stop the same person connecting with psql. The docs say so and point at the existing "Server Read-Only Is Not Safe Mode" section rather than implying more than it does.

No UI change yet. A managed control should render as managed rather than merely preset, and isManaged exists for exactly that, but Settings does not consume it in this PR.

Verification

• build PASS
• test ManagedPolicyResolverTests, ManagedPolicyReaderTests: 10 executed, 10 passed
• lint TablePro 0 violations

Tests use an isolated UserDefaults suite per case, so they need no profile installed and leave no state behind.

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated (UTC)
TablePro	🟢 Ready	View Preview	Aug 18, 2026, 8:42 PM

💡 Tip: Enable Workflows to automatically generate PRs for you.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.