Project Memory

Project Memory

The XHack Agent keeps a memory for each project: the target, the scope, decisions you have made, and what worked or failed. It reads and updates this memory automatically and loads it back at the start of every session, so you never have to re-explain your engagement. Memory is tied to each chat's working folder, so different jobs stay separate.

Set the working folder first

Memory follows the folder, so point the chat at the right one:

1. Above the input box, click Select Folder.
2. Choose the folder for this engagement.

From now on, everything the Agent remembers in this chat is stored against that folder. Open the same folder again later and the memory comes back with it.

Seed the memory

You do not fill in a form. You just tell the Agent. Type the facts that matter into the chat input box, for example:

"Remember for this project: scope is *.acme.com only. billing.acme.com is out of scope. Auth is a bearer token I'll paste each session. Priority is IDOR and auth bypass."

The Agent files these away on its own. As the job goes on, it keeps the memory up to date, noting confirmed findings, dead ends, and choices you made, without you asking.

How it recalls later

When you reopen the chat (or start a new session on the same working folder), the Agent auto-loads the memory before it does anything. You can pick up mid-engagement and it already knows the scope, the targets, and where things stood. You can always ask it what it remembers:

"What do you have in memory for this project so far?"

If something has changed, just tell it (for example, "Scope now includes shop.acme.com") and it updates the memory to match.

Related

• Sub-Agents: parallel workers share the project context
• Autonomous Agent: how the Agent plans across a session
• Prompting Guide: give the Agent clear scope to remember