Task decomposition strategies
Design task decomposition strategies for complex workflows.
This subtopic (1.6) sits in Agentic Architecture & Orchestration (D1) on Anthropic's Claude Certified Architect — Foundations (CCA-F) exam. The bank holds 20 practice questions here — 5 easy, 10 medium, and 5 hard — with 5 free to try, answers and explanations included. 3 of the free questions are below; the rest are in the practice stream.
What the exam tests here
- splitting large code reviews — per-file local passes plus cross-file integration pass
- dynamic decomposition for open-ended investigation tasks
- fixed sequential pipelines (prompt chaining) vs dynamic adaptive decomposition
- prompt chaining for predictable multi-aspect reviews
Free practice questions: Task decomposition strategies
A customer support agent must call get_customer before any order operation. Which decomposition approach enforces this sequence for open-ended investigation tasks?
Show answer & explanation
Correct answer: D. Programmatic prerequisite that blocks order tools until get_customer returns a verified ID
Programmatic prerequisites enforce tool ordering deterministically, unlike prompt or example-based approaches that rely on model compliance. 'System prompt instruction...' is probabilistic and can be skipped. 'Few-shot examples...' guide but do not guarantee ordering. 'A routing classifier...' controls tool availability, not sequencing within a workflow.
The synthesis subagent in your multi-agent research system must evaluate each source for credibility, relevance, and coverage before combining findings. Which task decomposition approach best ensures consistent, auditable evaluation across all three aspects?
Show answer & explanation
Correct answer: A. Run a separate chained prompt for each evaluation aspect, passing results forward as structured context to the next step.
Prompt chaining with one prompt per evaluation aspect produces consistent, auditable outputs because each step has a narrow scope and its result is visible before proceeding. 'Single prompt asking...all three aspects together' conflates concerns, making outputs harder to validate. 'Coordinator agent evaluate sources...' misplaces evaluation responsibility outside the synthesizer. 'Document analysis subagent to perform all three' conflates document parsing with synthesis-stage judgment.
The coordinator agent in a multi-agent research system must decide whether to invoke the web search subagent, document analysis subagent, or both, depending on whether a topic requires fresh data. Which decomposition strategy best handles this conditional routing?
Show answer & explanation
Correct answer: A. Dynamic adaptive decomposition where the coordinator evaluates topic requirements and selects subagents at runtime
Dynamic adaptive decomposition lets the coordinator assess each topic's needs and route selectively, avoiding unnecessary subagent calls. 'Fixed prompt chain always invoking all subagents' wastes resources when fresh web data is unnecessary. 'Static keyword matching at startup' cannot adapt to nuanced topic variations. 'Parallel invocation discarding outputs' wastes compute and synthesis capacity on irrelevant results.
2 more free questions on this subtopic in the practice stream, plus 15 in the full bank. Keep practicing →