IOS Data Persistence

iOS Data Persistence

Acknowledgement: Shared by Peter Bamuhigire, techguypeter.com, +256 784 464178.

Use When

• Designing, implementing, or reviewing iOS persistence with SwiftData, Core Data migration, UserDefaults, Keychain, file storage, offline sync, Core Spotlight semantic indexing, App Entity exposure, AI local context, or privacy-sensitive caches.
• The task mentions local storage, migrations, offline-first sync, search/indexing, Siri/App Entity data exposure, or AI cache retention.

Do Not Use When

• The task is general iOS implementation with no local data or cache concerns; use ios-development.
• The task is only platform capability routing; use ios-platform-capabilities.
• The task is security policy rather than persistence mechanics; use ios-security-and-rbac alongside this skill.

Required Inputs

• Data model, sensitivity classification, deployment target, sync/backend contract, offline requirements, account/tenant model, deletion rules, and whether data is exposed to Spotlight, Siri, widgets, or AI context.

Workflow

1. Load ios-development for Swift and project standards.
2. Choose storage: UserDefaults, Keychain, SwiftData/Core Data, FileManager, URLCache/NSCache, Core Spotlight, or an AI local-context cache.
3. Load references/ios-swiftdata.md for SwiftData details.
4. Load references/semantic-indexing-and-ai-caches-wwdc26.md when data is exposed to Siri, Spotlight, App Entities, widgets, or AI.
5. Produce a model spec, migration plan, cache policy, deletion policy, and test plan.

Quality Standards

• Persistence models must separate stored data, domain entities, DTOs, and search/intent projections when privacy or API shape differs.
• Secrets belong in Keychain or stronger platform storage, not UserDefaults, SwiftData, Spotlight, or logs.
• Offline data must have conflict, sync, stale-state, and deletion behavior.
• Indexed or AI-context data must have retention, invalidation, and user/account cleanup rules.

Anti-Patterns

• Exposing stored models directly as App Entities without reviewing fields.
• Keeping stale Spotlight entries after logout, role change, tenant switch, or deletion.
• Using local AI context as a hidden long-term memory without retention and deletion controls.
• Treating SwiftData migrations as a compile-time concern only.

Outputs

• Persistence model spec, storage decision, SwiftData migration plan, offline sync plan, search/index projection, AI cache policy, or review findings.

Evidence Produced

References

• references/ios-swiftdata.md for SwiftData @Model, relationships, model actors, migrations, CloudKit, and testing.
• references/semantic-indexing-and-ai-caches-wwdc26.md for Core Spotlight, App Entities, Siri semantic index privacy, local AI context, and cache invalidation rules.
• references/skill-deep-dive.md for UserDefaults, Keychain, repository pattern, file storage, URLCache/NSCache, offline-first architecture, and cross-skill references.

Storage Decision Guide

Inputs

Decision Rules

Degraded Mode

Without a runnable schema, provide the model delta, migration stages, backup/rollback path, and fixtures required for validation. Do not assert migration safety without testing old stores. If execution is unavailable, report migration safety as unverified.

Domain Anti-Patterns

• Storing tokens in UserDefaults. Fix: use an access-group-scoped Keychain item.
• Changing a persistent model without old-store fixtures. Fix: test each supported migration path.
• Caching private AI input indefinitely. Fix: define purpose-bound retention and deletion.
• Treating Spotlight as authoritative storage. Fix: rebuild the index from the source database.
• Resolving ledger conflicts by last-write-wins. Fix: reconcile against server versions.

Rule of thumb: simple flag/scalar = UserDefaults. Secret = Keychain. Relationships/querying = SwiftData. Binary blob = FileManager. Search/Siri/AI exposure = separate projection with explicit privacy review.