Agent Brain
The Agent Brain answers the question every reviewer asks: “What does this agent actually know and do?”
Where to find it
Section titled “Where to find it”| Place | Shows |
|---|---|
| Team page → agent ⚙ → 🧠 Atlas’s brain | The agent in this project |
| The agent’s engagement work item | The same, as a panel on the work item |
| The agent in the Agent Pool | Across projects: skills, connections, where it is hired, memories per project |
What it shows
Section titled “What it shows”
The agent sits in the middle; around it, everything that shapes its work:
- Skills — the approved instructions it follows, each linking to the skill work item.
- Memories — the lessons it has in this project, and where each was learned.
- Connections — the systems it may use here, with scope and policy: how many tools are read-only, ask first, or blocked, and how many you set to always allow.
- Polarion — the Polarion tools it may use.
- Project — who it reports to; the project instructions.
- Track record — tasks, runs, cost against budget, and first time right: the share of tasks you approved without sending back.
The exact instructions
Section titled “The exact instructions”▸ The exact instructions Atlas gets opens the full text the agent receives — built by the same code that runs it. Every sentence comes from a reviewed source: the persona, approved skills, the engagement, approved memories, and the platform’s fixed rules. Each run adds only its task: description, links and comments.