/guardx:scan — Run a security scan on a system prompt

Run a security scan on system prompts to identify vulnerabilities quickly.

Trigger

User invokes /guardx:scan or asks to "scan a system prompt for vulnerabilities".

Steps

  1. Ask the user: "Please provide the system prompt to scan — paste it directly or give me a file path."

  2. If the user provides a file path, use the Read tool to load the file contents.

  3. Call the scan_system_prompt MCP tool:

    Tool: scan_system_prompt
Arguments:
  systemPrompt: <the prompt text>
  mode: "dual"

  4. When the scan completes, immediately invoke the /guardx:interpret skill on the result to present structured findings to the user.

Notes

  • Default mode is dual (runs both extraction and injection testing).
  • If the user specifies a mode (extraction, injection, or dual), pass it through.
  • If the user specifies custom models or maxTurns, include those parameters.
  • Scans take 1–3 minutes depending on maxTurns. Let the user know before starting.