worldview-game-observation-gated-stalker
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [SAFE]: The skill details a rigorous design framework for implementing gameplay mechanics, focusing on state management and simulation ordering to ensure reliable behavior. It operates locally on user-provided assets and does not perform suspicious network or file operations.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external project files to inform its output. Ingestion points: The agent reads project repositories, scene data, and design briefs as described in SKILL.md. Boundary markers: The instructions tell the agent to scope its analysis only to user-authorized paths. Capability inventory: The skill can write implementation files and execute project-specific run/test commands. Sanitization: The methodology relies on scoping and user-verified intake notes rather than automated content sanitization.
- [COMMAND_EXECUTION]: The methodology involves the discovery and use of project-local build and testing tools. Evidence: SKILL.md instructs the agent to identify and run the project's "normal run and test commands".
Audit Metadata