new-user
Pass
Audited by Gen Agent Trust Hub on Mar 4, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection vulnerability surface through its argument handling.
- Ingestion points: User-provided
<email>and<password>arguments enter the agent context via theargument-hintand instructions. - Boundary markers: There are no delimiters or explicit instructions to the model to ignore embedded control characters or commands within the arguments.
- Capability inventory: The skill utilizes the
Bashtool to executenpx medusacommands. - Sanitization: No sanitization, escaping, or validation logic is specified for the user input before it is placed into the shell command string.
- [COMMAND_EXECUTION]: The skill constructs a shell command dynamically using
npx medusa user -e <email> -p <password>. Because the inputs are not sanitized, a malicious user could provide arguments containing shell metacharacters (e.g.,;,&&,|,`) to execute arbitrary commands on the host system.
Audit Metadata