at-scripting
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data in the form of 3D CAD models, which serves as a potential vector for indirect prompt injection attacks.
- Ingestion points: External data is ingested via
pxz.io.importSceneandpxz.io.importFiles, as documented inSKILL.mdandpxz-api-reference.md, and demonstrated inexamples/import-prepare-export.py. - Boundary markers: The provided code examples do not include explicit boundary markers or instructions to the LLM to ignore embedded data inside the processed files.
- Capability inventory: The skill provides full access to the Pixyz SDK, including the ability to write to the file system via
pxz.io.exportSceneandpxz.view.takeScreenshot, and to perform complex geometry and material transformations. - Sanitization: There is no evidence of sanitization or validation of the content within the imported model files before processing.
Audit Metadata