spatial-editor
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates a process where user-provided goals and asset descriptions are propagated to asset generation services. * Ingestion points: Untrusted content is ingested through the goal field in the start form, the objective in scene specifications, and the prompt field in model production plans (contracts.md). * Boundary markers: There is no instruction to use explicit delimiters or safety markers for the user-supplied text when it is interpolated into backend tool calls. * Capability inventory: The skill environment has broad capabilities to read project files, write USDA assets, and execute Editor-specific management tools (SKILL.md). * Sanitization: Technical material inputs (usdaText) are validated via validate_material_usda to confirm integrity before use (contracts.md).
- [COMMAND_EXECUTION]: The skill utilizes the pico-cli tool for synchronizing custom component declarations. * Consistent with the provided author context, pico-cli is a recognized vendor resource, and its execution is bounded to the intended project synchronization tasks described in the workflow (SKILL.md).
Audit Metadata