Skip to content

Questions & answers

Some decisions are yours alone: scope, money, taste, anything irreversible. The workflow’s rule for the agent is blunt:

When you need a decision only the owner can make, ask in the chat, right now. Wait for the answer; do not guess and continue. A wrong assumption costs more than a short pause.

The chat is where the question is asked — but chat scrolls away. So every answered question is recorded in the ticket, where the decision belongs:

## Questions
### Q1 · answered · 2026-08-06T10:00:00Z
Should the run history be persisted, or is session scope enough?
### A1 · bo · 2026-08-06T10:02:00Z
Session scope is enough.

Questions are numbered consecutively and numbers are never reused. The ticket is the record, not a mailbox: normally the Q/A block is written complete, after you answered.

If a run ends without an answer — you’re away, or the agent was started unattended — the question is written as open:

### Q2 · open · 2026-08-07T18:30:00Z
Which of the two paywall layouts should ship?

The ticket’s frontmatter gets open_question: Q2, the ticket stays in Doing, and the board shows it as waiting. You answer whenever you’re back — either by adding the ### A2 · bo · … block directly in the ticket file, or just in the next chat; the agent’s next run records it, clears the flag, and continues.

The effect, months later: every “why is it like this?” has a findable, timestamped answer sitting in the ticket that made the decision — not in a chat log nobody can search.