react-mcp
Pass
Audited by Gen Agent Trust Hub on Jun 12, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The documentation references downloading UI configurations from
https://r.assistant-ui.com/mcp-config.json. This is an official resource from the vendor used for component distribution. - [DATA_EXFILTRATION]: The skill describes using
McpLocalStoragefor persisting authentication tokens. It includes explicit security notes identifying that this method is XSS-exposed and recommends usingMcpCustomStoragewith secure backends for production environments. - [PROMPT_INJECTION]: The skill enables users to connect to third-party MCP servers, creating an inherent surface for indirect prompt injection if a server provides malicious tools.
- Ingestion points: User-defined URLs in the
McpAddFormPrimitivecomponent. - Boundary markers: MCP tools are namespaced by server ID (e.g.,
serverId__toolName). - Capability inventory: The skill exposes an API for executing remote tools from connected servers.
- Sanitization: The documentation does not specify sanitization of tool outputs, relying on host application logic and LLM safety layers.
Audit Metadata