editable-ui
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates a development workflow where the agent generates and manages source code files (
src/ui/*.tsx) that are intended to be parsed and modified by an external visual editor (Creator Hub). This establishes a trust relationship where the agent ingests code that can be influenced by external modifications, creating a surface for indirect prompt injection if the external tool's output is compromised or influenced by malicious inputs.\n - Ingestion points: Source files in the
src/ui/directory serve as a shared data format between the agent and the Creator Hub UI editor.\n - Boundary markers: The skill defines a strict structural 'contract' (permitted JSX elements, specific state object signatures) that acts as a boundary for the parser, though it does not provide logic-level sanitization.\n
- Capability inventory: The code produced by the skill has access to the full Decentraland SDK environment, including state management, event handling (onMouseDown, onChange), and the ability to register systems that execute logic every frame.\n
- Sanitization: The skill lacks explicit instructions for the agent to validate the security or integrity of code logic modified by the external editor before re-ingesting it into the project context.
Audit Metadata