How to use the Red team session template

Start a board from it
Use this template opens a new Freero board with the Red team session template on it (sign up free first). On a board you already have, open Templates in the toolbar and search for “Red team session”.
Follow the instruction on the board
“Invite people who didn’t build it. Write every attempt down, including ones that failed: they become test cases.”
Fill in each part
- Can it be talked into ignoring its instructions?
- Can it reveal data it shouldn’t?
- Can text inside a document steer it?
- Does it make things up when it doesn’t know?
- What do we fix first?
Bring your team and your agents
Invite people to the board or share a link, and work on it together live. The timer is in the top bar, private mode and voting in the board menu, and Claude or ChatGPT can join through Freero’s MCP server.

Work on it with an AI agent

Connect an assistant once and it works on the board with you, with your permissions: by default it reads and proposes, and people decide.

You ask: “Fill in the Red team session board from these notes.”
What happens It reads the board, then proposes sticky notes for it. Nothing changes until someone on the board accepts them.
You ask: “Group what we wrote into themes and list the actions.”
What happens It proposes clusters with a title each and task cards for the actions, labelled with its name and yours.

Agent permissions review
Decide what an AI agent may do on its own, what needs a person’s approval and what it must never do, action by action.
Prompt iteration log
Keep track of how a prompt evolves: each version, what changed, how it did on your test cases and whether to keep it.
AI risk review
Check an AI feature against the risks that matter before launch: how likely, how bad, what you do about it and who owns it.
Agent incident review
Look into a time an AI agent got something wrong: what happened, why it happened, why checks missed it, and what changes.
Model option comparison
Compare AI model or tool options for one job on quality, cost, speed, privacy and fit, using your own test cases.
AI feature spec
Specify an AI-powered feature: the user and job, inputs and outputs, how it can fail, the guardrails and how you’ll measure quality.