# Play Around in Prompt Hardening

This page focuses on **hands-on probing workflows** using Prompt Hardening. You can configure probes, execute prompts, and evaluate AI agent responses in real time.

## Editing Prompt Injection Probes

The Probe Editor allows you to configure prompt injection probes using a YAML-based definition.

You can use the Probe Editor to:

* Define prompt injection and adversarial attack patterns.
* Configure detection rules for unsafe or non-compliant responses.
* Assign severity levels to classify identified risks.

The YAML-based configuration enables iterative probing without redeploying AI agents, which helps maintain rapid feedback during security validation.

## Probing Prompts

The Playground provides an interactive environment for executing prompt hardening probes against AI agents.

#### Navigating to the Playground

You can access the Playground using the following navigation path:

**Agentic Security → Prompt Hardening → Playground**

<figure><img src="/files/OWfdWV9h3hVkKEKEp5RH" alt="" width="563"><figcaption></figcaption></figure>

The Playground supports real-time prompt execution and response inspection.

### Playground Input Areas

Each Playground session separates trusted instructions from adversarial input to support controlled probing.

{% tabs %}
{% tab title="System Prompts" %}
System Prompts define the baseline instructions provided to the AI agent. System Prompts represent the trusted control layer under probe.

<figure><img src="/files/DTafvj10iH2a81iTO15I" alt="" width="563"><figcaption></figcaption></figure>
{% endtab %}

{% tab title="User Input" %}
User Input allows injection of malicious or adversarial content. You can manually enter prompts or generate them automatically.

The **Auto-generate prompt** option creates adversarial input using predefined attack templates.

<figure><img src="/files/iLJ35sl2ucAR2wKc2xBZ" alt="" width="563"><figcaption></figcaption></figure>
{% endtab %}

{% tab title="Auto Generate Prompt" %}
Auto Generate Prompt uses Akto-defined attack patterns to create custom malicious prompts. This option supports rapid scanning without manual prompt design.
{% endtab %}
{% endtabs %}

### Executing Prompt Probes

After configuring System Prompts and User Input, you can start scanning by selecting **Probe System Prompt**.

The Playground sends the combined prompt to the AI agent and captures the response for analysis.

### Reviewing Agent Responses

After execution, the Playground displays the AI agent response.

Each result includes:

* The complete agent output.
* A safety status indicating whether the response complied with expected controls.
* A vulnerability indication when unsafe behavior or policy violations are detected.

<figure><img src="/files/pg4gSuUqRQHucn5PpoSn" alt="" width="563"><figcaption></figcaption></figure>

Response visibility helps you quickly identify prompt injection risks and unsafe AI agent behaviour during scanning.


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://ai-security-docs.akto.io/akto-argus-agentic-ai-security-for-homegrown-ai/prompt-hardening/play-around-in-prompt-hardening.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.