spatial-computing
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references standard technical resources, including the SwiftTerm project on GitHub and official Apple Developer documentation. These are reputable sources and do not escalate the security risk.
- [INDIRECT_PROMPT_INJECTION]: The skill acts as a coordinator for specialized spatial computing tasks, creating a surface for indirect injection via user requests. 1. Ingestion points: User objectives and deliverable descriptions in SKILL.md. 2. Boundary markers: The skill uses a structured 'Procedure' to scope interaction but lacks explicit boundary delimiters for references. 3. Capability inventory: The skill is composed entirely of markdown documentation and contains no active executable scripts, subprocess calls, or network operations. 4. Sanitization: No sanitization logic is implemented in the instructions.
- [SAFE]: The provided code snippets for Metal rendering, visionOS interfaces, and spatial interaction systems are purely instructional and follow standard platform best practices.
Audit Metadata