agui
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill uses
npx symbiote-aguito execute the vendor's toolset from the npm registry and downloads verifiedcloudflaredbinaries from official Cloudflare releases for internet sharing capabilities.\n- [REMOTE_CODE_EXECUTION]: Remote execution is present via thenpxmechanism and binary downloads; however, these are restricted to well-known services and the verified vendor repository.\n- [COMMAND_EXECUTION]: The skill defines a feedback mechanism that allows the agent to execute local commands for external synchronization, which is managed via private registration files.\n- [DYNAMIC_EXECUTION]: The framework allows for the dynamic execution of JavaScript modules within HTML fragments via Blob URLs for interactive UI components.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted HTML and Markdown data, creating a potential injection surface.\n - Ingestion points: Data enters through HTML insertion, Markdown rendering, and JSON state updates (e.g., in
SKILL.md).\n - Boundary markers: The instructions mandate clear labeling and structured data envelopes for all external content.\n
- Capability inventory: The skill includes file system operations via the CLI, network tunneling, and local command execution.\n
- Sanitization: The skill commands the agent to disable untrusted document extensions and explicitly forbids treating document content as executable code or authorization instructions.
Audit Metadata