designer-session
Pass
Audited by Gen Agent Trust Hub on Sep 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The
export_mapstool (implemented inscripts/export_maps.py) invokes the officialsbsrenderexecutable. The skill documentation states this is a bounded operation using the official CLI discovered by the host environment. - [INDIRECT_PROMPT_INJECTION]: The skill processes data from Substance 3D Designer graphs, which acts as a potential surface for indirect prompt injection.
- Ingestion points:
scripts/export_graph_state.pyreads node identifiers, parameter values, and connection topology from the active Designer graph. - Boundary markers: The extracted graph state is returned to the agent context as structured JSON without specific boundary markers or 'ignore' instructions for the LLM.
- Capability inventory: The skill possesses file system write capabilities (
scripts/save_package_as.py,scripts/export_maps.py,scripts/export_sbsar.py) and graph mutation capabilities. - Sanitization: Identifiers provided as input to authoring tools are normalized via regex in
_normalize_identifier, though data read from existing graphs is returned to the agent as-is.
Audit Metadata