Skip to main content

​
Purpose

The mindgard recon guardrail command lets you detect whether a target system has a guardrail in place before committing to deeper testing. You can optionally extend this recon to attempt identify the type of guardrail through fingerprinting. Based on the findings from this tool, you can then customise your tests to attempt bypasses and/or adjust payload strategies for deeper assessment.

​
Usage

After you have logged into your Mindgard account in the CLI, run: 
mindgard recon guardrail --config-file {YOUR_CONFIG.toml}
Arguments --config-file (required) : Path to your TOML configuration file. More information on configuration files can be found here. Fingerprinting If mindgard recon guardrail detects the presence of a guardrail, you will be offered the ability to perform fingerprinting to attempt to identify exactly which guardrail is in place e.g. Protect AI v2. Selecting y when prompted will start fingerprinting recon. Be aware that this will run ~120 prompts against your target. Results At the end of the recon test you will be given a link to its results page. The results page will present guardrail information, prompt/response details, and recommendations for further action.

​
When to use it

  • Early in an engagement to determine if guardrails are present.
  • To plan follow-up tests more efficiently based on the presence or absence of guardrails.

​
Access

This feature is only available to approved organizations or users on request. Contact Mindgard support to enable it.