Architect
The Architect agent designs a solution that fits your constraints and existing system. It breaks work into pieces the team and the product can execute.
When to use Architect
- You have agreed requirements and need a technical approach.
- You want an ADR for traceability.
- You need a sensible decomposition into implementation steps.
Typical outputs
| Output | Purpose |
|---|---|
| ADR | Records the chosen approach and tradeoffs |
| Stack | Notes on frameworks, services, and tooling |
| Suggested tasks | Proposed implementation slices for review |
| Task spec | Detailed instructions copied to child coder work during epic execution |
Attach the Analyst artifact when you open the Architect step so decisions align with requirements.
Suggested tasks in the current product
Suggested tasks are proposals, not immediate execution. In the current product you can:
- dismiss items that are out of scope
- restore previously dismissed items
- use Implement all to turn the remaining tasks into child coder work with task specs
This keeps the board under human control while still benefiting from AI decomposition.
After Architect
- Review the ADR and task list while the card is in wait_for_review.
- Start coder execution from the Architect result using Implement all.
- For larger initiatives, use Epic mode to batch subtasks under one parent.