browserenginekit
Pass
Audited by Gen Agent Trust Hub on Sep 11, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [SAFE]: The skill provides technical guidance for using the BrowserEngineKit framework on iOS, focusing on legitimate development patterns and security features. Each component (Web Content, Networking, Rendering) follows the principle of least privilege through process isolation and system-brokered communication.
- [EXTERNAL_DOWNLOADS]: The documentation links reference the sosumi.ai domain for framework specifications. These are static references and do not involve runtime code execution, untrusted package installations, or piped shell commands.
- [INDIRECT_PROMPT_INJECTION]: The architecture handles untrusted web content as its primary function and defines a clear security posture for this attack surface.
- Ingestion points: Web content extensions ingest HTML, CSS, and JavaScript from untrusted remote servers.
- Boundary markers: The architecture mandates process isolation via XPC and the application of a restricted sandbox using applyRestrictedSandbox(revision: .revision2).
- Capability inventory: The skill defines process management (WebContentProcess, NetworkingProcess, RenderingProcess), network operations, file access via security-scoped bookmarks, and JIT compilation.
- Sanitization: Integration with BEWebContentFilter is provided to evaluate and restrict URLs based on system-level policies and parental controls.
- [DYNAMIC_EXECUTION]: The skill documents the secure use of JIT compilation for JavaScript engines. This requires specific entitlements (allow-jit and extended-virtual-addressing) and utilizes memory protection tags (BE_JIT_WRITE_PROTECT_TAG) to control JIT memory page permissions, following standard security practices for mobile browsers.
Audit Metadata