ag-ui-protocol

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSNO_CODE
Full Analysis
  • [No Code] (SAFE): The skill consists exclusively of Markdown documentation and does not ship with any executable scripts, binaries, or automated configuration files.
  • [Indirect Prompt Injection] (LOW): The AG-UI protocol defines an interaction model that processes untrusted data from users and external tools, which constitutes an attack surface for indirect prompt injection. * Ingestion points: User message content and tool results (see docs/2025-11-27/concepts/messages.md and docs/2025-11-27/concepts/tools.md). * Boundary markers: The protocol uses structured event types and roles (user, assistant, system) as boundaries. * Capability inventory: The skill metadata in SKILL.md allows the use of Bash and Edit tools. * Sanitization: The provided documentation does not include explicit sanitization or filtering logic for ingested content.
  • [External Downloads] (LOW): Documentation and example setup instructions reference resources from the ag-ui-protocol GitHub organization and associated npm packages. While these sources appear legitimate for the protocol's purpose, they are not on the predefined list of trusted external sources.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:05 PM