game-dev

Pass

Audited by Gen Agent Trust Hub on Oct 4, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute local commands for project verification, including running type checks, unit tests, and production builds to ensure the game is playable.
  • [EXTERNAL_DOWNLOADS]: The skill interacts with the SandBase service using a suite of tools (sandbase_run, sandbase_run_get) to generate and retrieve game assets like textures and sprites. The instructions mandate clear communication of costs and explicit user approval before any generation calls are made.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process a 'game brief' from the user, which acts as a primary ingestion point for untrusted data. This data influences code generation and asset prompts. The skill mitigates risks by focusing on local implementation and requiring human-in-the-loop confirmation for external asset generation.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 4, 2026, 11:45 AM