sogni-worlds
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches the 'Sogni Worlds Kit' from the vendor's official GitHub repository ('Sogni-AI/sogni-worlds-kit') and installs dependencies via 'npm install'.
- [COMMAND_EXECUTION]: Executes shell commands using 'node', 'git', and 'npx' to manage the world-building lifecycle, including rendering, building, and deploying via Cloudflare's 'wrangler' CLI.
- [INDIRECT_PROMPT_INJECTION]: Processes data from external files ('AGENTS.md', 'lessons.md') and user-provided photos which could contain instructions.
- Ingestion points: Kit-provided markdown files ('AGENTS.md'), local lesson logs ('~/.config/sogni/lessons.md'), and user assets.
- Boundary markers: None identified; the agent is instructed to follow instructions contained within these files.
- Capability inventory: Includes network operations ('git', 'wrangler'), file system access, and media processing ('ffmpeg').
- Sanitization: Not explicitly mentioned in the skill instructions.
Audit Metadata