app-intents
Pass
Audited by Gen Agent Trust Hub on Aug 23, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill incorporates robust security principles for handling sensitive data within the iOS ecosystem. It explicitly cautions against including private information such as account balances, tokens, or personal account numbers in shared system surfaces like Spotlight search indexes (IndexedEntity) or widget configurations.
- [SAFE]: It promotes the use of fine-grained security controls provided by the App Intents framework, such as 'authenticationPolicy' to ensure the device is unlocked and 'requestConfirmation' to prevent accidental or malicious execution of sensitive actions.
- [INDIRECT_PROMPT_INJECTION]: The skill defines a standard surface for processing user-supplied data through intent parameters and entity queries.
- Ingestion points: Data enters the system via '@Parameter' properties in 'AppIntent' and query strings in 'EntityStringQuery' (found in SKILL.md and references/entities-queries.md).
- Boundary markers: The framework provides natural boundaries through typed parameters and the use of 'LocalizedStringResource' for system dialogs.
- Capability inventory: The skill's 'perform()' function allows execution of application business logic, such as 'TransferService.send()', which represents the primary capability associated with the ingested data.
- Sanitization: The skill mitigates risks by requiring explicit confirmation prompts and local device authentication policies for any sensitive capabilities, ensuring a human-in-the-loop for critical actions.
Audit Metadata