BuildMidData / MLAuto-graded
Route each subtask to the right agent
The unglamorous coordination details decide this one: the right handler, results in order, no crash on the unexpected.Real AI systems rarely use one model for everything. You decompose a job into subtasks, hand each to the specialist best suited to it, and stitch the answers back together. The piece deciding who handles what, and reassembling the results, is the orchestrator — the backbone of any multi-agent system. The one inorchestrator.py is broken. Make it route each subtask correctly, keep results in order and attached to the right subtask, and handle an unknown task type properly both with and without a default agent.30 min
python
You're done when
- The orchestrator runs a mixed batch of subtasks without crashing.
- Each subtask reaches the specialist declared for its type.
- Results come back in order, each carrying the id of the subtask it belongs to.
- There is no cross-talk — one subtask's result never lands on another.
- With a default agent configured, an unknown task type falls through to it.
- With no default configured, an unknown type produces an error result instead of crashing, and is never misrouted to an unrelated specialist.
- Known task types are still routed correctly in that no-default configuration.
What it measures
Agent engineeringBuilding the dispatch layer that lets specialist agents be composed into one system.
Context engineeringKeeping each subtask's context and result bound together as work fans out and back in.
Handling the unrouteableDeciding what an unknown task type should do in each configuration, rather than defaulting to whichever agent is first.
Your environment
Terminal + editorA full shell in the browser
Claude CodePre-authenticated, no key setup
Isolated microVMYours alone, wiped afterwards
Start this lab
30 min
Start labYou'll sign in to Anthropos to start. The lab opens in a private cloud workspace with Claude Code ready to use.AI Labs is part of Anthropos Workforce. Not on your plan yet? Book a demo