Lane context is the information ADE gives an agent so it starts in the right place. It is intentionally small and local to the work.
What context includes
- Lane name, branch, base branch, and worktree.
- The current chat or terminal session.
- The task prompt and any explicit instructions.
- Artifacts you attach, such as proof, logs, screenshots, or linked issues.
- PR and Linear context when the lane is linked.
What agents still need to do
Agents should inspect the repository directly when they need code, architecture, tests, or product details. ADE does not rely on a giant generated summary as the source of truth.
Improve a lane’s context
Name the lane well
A clear lane name gives humans and agents the first clue.
Write a focused prompt
Say what should change, where to look, and how to verify it.
Attach useful artifacts
Add screenshots, logs, Linear issues, or PR comments when they matter.
Keep follow-up in the same lane
Continue the same chat or start a new lane-scoped chat when the work changes direction.
Good context is specific, not large. A failing test name and expected behavior often help more than a long background essay.
