beui
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches a live component registry in JSON format from
https://beui.dev/r/registry.jsonusingcurl. This is the standard mechanism used to discover available components within the library's ecosystem.- [COMMAND_EXECUTION]: The skill provides instructions for the agent to execute shell commands vianpx,pnpm, orbunto interface with theshadcnCLI. These commands are used to view and add components to the user's local project.- [REMOTE_CODE_EXECUTION]: The use ofnpx shadcn@latest add @beui/<slug>downloads external component code from the beUI registry and integrates it into the project files. This is a functional requirement for component libraries following the shadcn distribution model.- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from an external JSON registry to determine which components to install. Ingestion points: Registry metadata (names and slugs) is fetched frombeui.dev. Boundary markers: None; the agent is instructed to map user intent directly to the slugs found in the JSON. Capability inventory: The skill has shell execution capabilities vianpxandcurl. Sanitization: No explicit validation or sanitization of the remote registry content is performed before the data is used in shell commands. This represents a vulnerability surface typical for registry-based discovery tools.
Audit Metadata