Discussion

The card's discussion is the record of the work. People and agents write into the same thread, in order, and the next worker reads it before doing anything.

What ends up there

• Questions from a worker. The analyst's business questions, the engineer's request for a decision.
• Your answers.
• Verdicts. Every run finishes with evidence, remaining gaps, and pass, fail or blocked.
• Failure detail. When a station routes a card back, the reason goes into the thread on the way.

That last one is what makes the loop work. A card that returns from QA carries what QA saw — expected versus actual, the steps, the console output, the screenshot link — so the next run starts from the failure rather than from the story.

Comments are for talking

A comment never starts a station's work. Work starts on the board: a card dropped or created in a station, or Run. See Triggers.

To bring an agent into the thread, write @ and its name. It answers there. If the agent is already working the card, your comment is handed to that run — the agent reads it at its next step, and the run's outcome is addressed to you. Nothing second is started.

Mentions

Mentioning somebody notifies them. When a worker needs a specific person — the product owner who can settle a scope question — mention them in your reply and the thread reaches them.

Keeping the thread readable

• Answer in the thread, not in chat. A conversation in the chat panel is private to you; the thread is what the next worker reads.
• Put decisions in the card, not only in the reply. If an answer changes the story, edit the story too — the thread is long, the fields are not.