search-registry-items
Pass
Audited by Gen Agent Trust Hub on Aug 3, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill guides the agent to execute
npx shadcn@latest addshell commands. The command arguments are dynamically populated from the results of thesearch_registry_itemsMCP tool. This behavior is the intended functionality for component installation. - [EXTERNAL_DOWNLOADS]: The skill interacts with the Shoogle MCP server (
mcp.shoogle.dev) to fetch registry data. It also references official documentation hosted on the shadcn-ui GitHub repository, which is a recognized source for this ecosystem. - [PROMPT_INJECTION]: Indirect Prompt Injection Surface (Category 8):
- Ingestion points: The
addCommandArgumentfield returned by thesearch_registry_itemstool output, as documented inreferences/TOOL-REFERENCE.md. - Boundary markers: None; the remote data is interpolated directly into a shell command string.
- Capability inventory: Shell command execution capability via
npx. - Sanitization: None; the instructions explicitly state to use the argument verbatim. While this represents a surface for injection if the remote registry service were compromised, it is consistent with the primary purpose of a developer integration tool.
Audit Metadata