watermelon-ui
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill connects to an external, public MCP server at
https://mcp.watermelon.sh/mcpto fetch UI component metadata, source code, and dependencies. - [COMMAND_EXECUTION]: The workflow involves executing installation commands (such as shadcn-ui) provided by the external service to integrate components into the local project environment.
- [INDIRECT_PROMPT_INJECTION]: The skill processes source code and data from an external repository, which introduces a surface for instructions potentially embedded in the retrieved content.
- Ingestion points: Components, blocks, and templates retrieved via the
get_componentandcompose_pagetools. - Boundary markers: The instructions do not define specific delimiters to separate external code from agent instructions.
- Capability: The agent has the capability to write the retrieved source code to the local file system and run associated installation commands.
- Sanitization: The skill does not specify procedures for sanitizing or validating the code returned from the external server.
Audit Metadata