Two agents run in parallel and keep changing the same files. What should you configure first?
Why B is the answer
Isolation so each works in its own branch or workspace
Isolation for parallel execution prevents overlapping changes, and conflict detection handles what remains.
In detail
- Isolation gives each agent its own branch or workspace.
- Work cannot collide while it runs.
- Split the work so that files overlap as little as possible.
Use case
An illustrative scenario, not a real incident.
Two agents keep editing the same config. With its own branch for each, they work without collisions, and any overlap is resolved at merge.
See it in different ways
flowchart TD A["Coordinator"] --> B["Agent 1 on branch a"] A --> C["Agent 2 on branch b"] B --> D["Pull request a"] C --> E["Pull request b"] D --> F["Review and merge in order"] E --> F
stateDiagram-v2 [*] --> CheckDependencies CheckDependencies --> Sequential: one step needs another CheckDependencies --> Parallel: work is independent Sequential --> Handoff: clear artifact Parallel --> Isolated: own branch or workspace Handoff --> [*] Isolated --> [*]
mindmap
root(("Parallel agents"))
n1["Agent A"]
n2["Own branch"]
n3["Own workspace"]
n4["Agent B"]
n5["Own branch"]
n6["Own workspace"]Key point: Isolate parallel agents.
Why the other options are wrong
- A. Longer timeouts so each agent can finish
- More time gives more chance of conflict.
- C. A shared chat channel where they post their plans
- Posting plans does not prevent collisions.
- D. Faster models so the overlap window is shorter
- A shorter overlap is still an overlap.