cloudkit

Pass

Audited by Gen Agent Trust Hub on Sep 11, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill describes patterns for fetching and processing records from CloudKit databases, which establishes an attack surface for indirect prompt injection.\n
  • Ingestion points: Untrusted data is ingested through record change blocks such as fetchedRecordZoneChanges in the SyncManager class (SKILL.md) and recordWasChangedBlock in IncrementalSyncManager (references/cloudkit-patterns.md).\n
  • Boundary markers: The provided implementation examples lack explicit delimiters or instructions to treat remote record content as untrusted data.\n
  • Capability inventory: The skill includes capabilities to modify files on the local system via FileManager.default.moveItem (references/cloudkit-patterns.md) and to perform writes to the CloudKit database using privateDB.save (SKILL.md).\n
  • Sanitization: There is no evidence of data sanitization, schema validation, or escaping of external record values before they are processed or persisted.\n- [EXTERNAL_DOWNLOADS]: The documentation contains multiple hyperlinks to sosumi.ai. While this is a known community mirror for Apple developer documentation, it is a third-party domain that is not part of the official vendor infrastructure.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 11, 2026, 11:14 AM
Security Audit — agent-trust-hub — cloudkit