tinyworld-asset-editing
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and process untrusted external data which could potentially contain malicious instructions.
- Ingestion points: Untrusted data enters the agent context via chat-attached model imports (window.__tinyworldAgentDropAttachments) and local file imports (GLB, GLTF, OBJ, and voxel JSON) described in SKILL.md.
- Boundary markers: The instructions lack explicit boundary markers or delimiters when interpolating asset summaries or attachment IDs into generator prompts.
- Capability inventory: The skill has the capability to write to the world state using setCell() and applyState(), and it can persist data in localStorage and IndexedDB.
- Sanitization: The skill mitigates risks by requiring that generated world changes be staged as a holographic diff overlay for manual user approval ('Apply / Regenerate / Discard') before committing the state.
Audit Metadata