# Task decomposition: chains vs dynamic plans

*Claude Certified Architect - Foundations Study Guide · Agentic Architecture & Orchestration · Last updated October 22, 2026 · https://www.certifyforce.com/study-guide/claude-certified-architect-foundations/agentic-architecture-and-orchestration/task-decomposition-chains-vs-dynamic-plans*

How to split complex work for Claude: fixed prompt chains for predictable multi-aspect reviews, parallel sections for independent checks, per-file passes plus a cross-file integration pass for large reviews, and adaptive plans that map, prioritize and re-plan for open-ended tasks.

- Certification: Claude Certified Architect - Foundations
- Exam section: Agentic Architecture & Orchestration
- Study guide: https://www.certifyforce.com/study-guide/claude-certified-architect-foundations

**Official exam objective:**

> Design task decomposition strategies for complex workflows

## What the exam expects

- Choose a fixed prompt chain for predictable multi-aspect work and dynamic, adaptive decomposition for open-ended investigation, from signals in the requirement.
- Run independent checks in parallel and sequence steps that depend on another step's output (the schema before its implementers).
- Split a large code review into per-file local passes plus a separate cross-file integration pass to avoid attention dilution.
- Decompose an open-ended task ("add comprehensive tests to a legacy codebase") by mapping structure, finding high-impact areas and writing a prioritized plan that adapts as dependencies appear.
- Run a large mechanical change as a task list, one `claude -p` call per item, piloted on a few items first.

Key terms: **prompt chaining** (each call processes the previous call's output); **sectioning** (independent subtasks run in parallel and aggregated); **orchestrator-workers** (a central model breaks the task down at run time); **attention dilution** (quality falling as one pass covers too much); **adaptive plan** (a plan rewritten from intermediate findings).

## Key facts

| Fact | Detail | Source |
|---|---|---|
| Prompt chaining | Each call processes the previous output; for fixed subtasks | [Building effective agents](https://www.anthropic.com/research/building-effective-agents) |
| Chain trade-off | More latency for higher accuracy; gates on intermediate steps | [Building effective agents](https://www.anthropic.com/research/building-effective-agents) |
| When to chain explicitly today | To inspect intermediate outputs or enforce a pipeline structure | [Prompting best practices](https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices) |
| Sectioning | Independent subtasks in parallel, aggregated programmatically | [Building effective agents](https://www.anthropic.com/research/building-effective-agents) |
| Orchestrator-workers | Subtasks determined at run time; e.g. file count depends on the task | [Building effective agents](https://www.anthropic.com/research/building-effective-agents) |
| Large reviews | Per-file local passes plus a cross-file integration pass | [Exam Guide](https://everpath-course-content.s3-accelerate.amazonaws.com/instructor%2F6nizmqk8tpzpfjvt6qmmav7rh%2Fpublic%2F1783542750%2FClaude+Certified+Architect+%E2%80%93+Foundations+Exam+Guide.pdf) |
| Research is path-dependent | Lead agent decides whether more research is needed and spawns more subagents | [Multi-agent research system](https://www.anthropic.com/research/multi-agent-research-system) |
| Search strategy | Short, broad queries first, then narrow | [Multi-agent research system](https://www.anthropic.com/research/multi-agent-research-system) |
| Fan-out pilot | Refine the prompt on the first 2-3 files, then run the full set | [Best practices](https://code.claude.com/docs/en/best-practices) |

## Check yourself

- When should you choose orchestrator-style dynamic decomposition over a prompt chain?
- Why does a separate cross-file integration pass catch bugs that per-file passes miss?
- What do you gain and lose by turning one generation call into a three-step chain?
- Why should research subagents start with short, broad queries?

## Sources

- [Building effective AI agents](https://www.anthropic.com/research/building-effective-agents)
- [How we built our multi-agent research system](https://www.anthropic.com/research/multi-agent-research-system)
- [Prompting best practices](https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices)
- [Best practices for Claude Code](https://code.claude.com/docs/en/best-practices)
- [Common workflows](https://code.claude.com/docs/en/common-workflows)
- [Claude Certified Architect – Foundations Exam Guide v1.0](https://everpath-course-content.s3-accelerate.amazonaws.com/instructor%2F6nizmqk8tpzpfjvt6qmmav7rh%2Fpublic%2F1783542750%2FClaude+Certified+Architect+%E2%80%93+Foundations+Exam+Guide.pdf)

## About the full unit

The "Core concepts", "Exam traps" and "Worked scenario" sections of this unit, its 11 linked practice questions, and unlimited timed mock exams are included with the Claude Certified Architect - Foundations pack (USD $19): https://www.certifyforce.com/practice-exams