You should enable mask_inpainting and supply both the input image and the mask to meet all the criteria for your Microsoft Foundry image-editing workflow.
By utilizing Microsoft Foundry's built-in Image Generation Tool parameters, configuring the workflow this way ensures the desired edits are perfectly executed.
Workflow Configuration Requirements
Mask-Based Object Removal: Passing the mask parameter explicitly flags the exact background object regions targeted for removal, replacing them seamlessly.
Preserving Style and Lighting: Enabling mask_inpainting prompts the underlying built-in model (such as gpt-image-2) to inherit and maintain the exact lighting, textures, and style of the surrounding unmasked environment.
Built-In Controls Only: This is entirely handled natively via the Foundry Agent Service API variables without deploying a single line of custom code or third-party model checkpoints.
Targeted Area Enforcement: The system relies on the provided mask array to ensure pixels outside the marked coordinates are kept completely untouched and protected from VAE degradation.
Reference:
https://learn.microsoft.com/en-us/azure/foundry/agents/how-to/tools/image-generation

Explanation:
Prompt shields action: Set action to block.
Additional mitigation: Enable Spotlighting.
The correct configuration is to set the Prompt Shields document attack action to block and enable Spotlighting . Prompt Shields in Microsoft Foundry are designed to detect attempts to manipulate model behavior through adversarial input. Microsoft distinguishes document attacks from direct user prompt attacks:
document attacks are malicious instructions embedded in third-party content such as documents, webpages, emails, or other externally supplied material. In this scenario, the embedded text inside uploaded screenshots is third-party content and can attempt to override the agent's instructions. Setting the action to block prevents detected document-attack content from being processed normally, which is required because the goal is prevention rather than passive logging or annotation.
Spotlighting is the additional mitigation because it marks or transforms document content so the model treats it as lower trust than system and user instructions. Microsoft's Foundry guidance describes Spotlighting as a Prompt Shields subfeature that helps protect against indirect or embedded document attacks by tagging input documents with special formatting to indicate lower trust. A custom blocklist is insufficient for unknown attacks, and OCR alone only extracts the malicious text; it does not mitigate prompt injection. Reference topics: Prompt Shields, document attacks, guardrails, Spotlighting, multimodal safety, and prompt injection defense.

To best manage security and centrally handle credentials across multiple agents, you should add a connection to the Azure AI Search resource at the Azure AI Foundry project level.
Why This Works
Central Hub: The project acts as the single security perimeter for all your agents.
Credential Masking: Agents inherit access without hardcoding secrets, API keys, or connection strings in their code.
Identity Management: It allows you to leverage Microsoft Entra ID (formerly Azure AD) for role- based access control (RBAC).
How to Implement It
1. Navigate to your Azure AI Foundry portal.
2. Select your specific project from the dashboard.
3. Open the "Management Center" or "Project settings" tab.
4. Click on "Connected resources" or "Connections".
5. Add the Azure AI Search resource.
6. Choose Entra ID (managed identity) over API keys for maximum security.
Reference:
https://partner.microsoft.com/en-us/blog/article/azure-updates-december-2025

To monitor and control access to a custom agent in Azure, you must first create a Microsoft Foundry project. Once the project is created, you register your custom agent within it to enable management capabilities such as access control and activity monitoring.
Reference:
https://learn.microsoft.com/en-us/azure/foundry/control-plane/register-custom-agent

E xtractive summarization within the Azure AI Language service is the appropriate feature for the application's requirements.
This feature can be configured to:
Return exactly three sentences by using the sentenceCount parameter (which defaults to 3). The supported range is from 1 to 20 sentences.
Preserve the original sentence order by specifying the sortBy parameter as Offset, which is the default behavior. The Offset value corresponds to the character position of the extracted sentences in the original document.
Reference:
https://docs.azure.cn/en-us/ai-services/language-service/summarization/how-to/text-summarization