vite
Warn
Audited by Gen Agent Trust Hub on Mar 17, 2026
Risk Level: MEDIUMPROMPT_INJECTIONCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [PROMPT_INJECTION]: The skill metadata identifies 'Anthony Fu' as the author, which conflicts with the technical submission author 'ggglhhh', representing a misleading metadata pattern.
- [PROMPT_INJECTION]: The skill is designed to process untrusted local data such as project source files and environment variables, creating an indirect prompt injection surface.
- [COMMAND_EXECUTION]: Documentation includes CLI commands and JavaScript APIs for building and serving projects, which involve executing local processes and scripts.
- [REMOTE_CODE_EXECUTION]: The skill describes the 'ssrLoadModule' API and plugin hooks that facilitate the dynamic loading and execution of code modules within the runtime environment.
- [DATA_EXFILTRATION]: Code examples demonstrate the use of 'loadEnv' with empty prefix filters, which could lead to sensitive environment variables being loaded into memory and potentially exposed.
Audit Metadata