Creating a Blank Workspace

This article shows how to create a Blank Workspace — a free-form, chat-driven pentest engagement where you describe the task in plain English and let the AI agent plan the methodology.

Prerequisites

Before starting a Blank Workspace, ensure the following:

• You have a clear description of what you want the agent to investigate.
• You have any reference files ready (scope docs, prior pentest reports, screenshots, .pdf, .docx, .txt, .md, .html, or .csv files).
• You have any credentials the agent might need saved under Settings → Credentials (you can also attach them later in chat).
• Your organization has sufficient AI credits.

Steps to start a Blank Workspace

Step 1: Pick the Blank Workspace template

• Open AI → Workspaces and click New Workspace.
  
  
  
• On the Template step, click the Blank Workspace card (the first card, tagged Custom).

Step 2: Describe the task

• A composer appears with a free-form textarea.
• Describe what you want the agent to do, in plain English. Example:

Investigate https://app.acme.com/api/v3/checkout for IDOR.
We've seen one report that order_id is sequentially numbered.
I've attached the prior pentest report as reference.
Confirm exploitability and produce a write-up.

Step 3: Attach reference files

Use Add files to drop in scope docs, screenshots, prior reports, or any supporting material. Supported formats include .pdf, .docx, .txt, .md, .html, and .csv.

Step 4: Configure runtime

• Model: Pick the LLM tier — Standard, Advanced, or Lite.
• Browser: Cloud Browser (default) or Connected Browser.
• Shell: Optional. Attach if your task requires internal-network access.
• Auto-approve: Toggle on to let the agent run autonomously without pausing for confirmations.

Step 5: Launch

• Click Create & Launch.
• You're routed straight into the AI Chat for the new Workspace. The agent reads your description, plans the steps, and begins.
• Continue to guide the agent from chat as needed.

Where Can This Be Used?

Blank Workspaces are best for:

• Custom investigations that don't fit a standard template (e.g., "validate this CVE PoC against our staging server").
• Targeted exploit confirmation when a single bug needs deeper validation.
• Exploratory work where you want the AI to design the methodology itself.
• Chat-first engagements where you prefer a conversational workflow over a structured assessment.

Note: A Blank Workspace can always be promoted into a more structured assessment by asking the agent in chat — for example, "Save this as an Engagement and produce a formal Web App Pentest report."

Troubleshooting Tips

• If the agent immediately asks for clarifying questions, your description was likely too vague — re-state the goal, target, and what success looks like.
• If the agent can't reach the target, attach a Connected Browser or Shell depending on whether the target is a web app or an internal-network service.
• If your reference file contains sensitive credentials, scrub them before attaching — files become part of the Workspace context.