Home › Study guides › CCAR-F › Domain 1

CCAR-F · Domain 1 of 5 · 7 lessons · about 143 min

Domain 1: Agentic Architecture & Orchestration

How a Claude agent runs, how a coordinator hands work to subagents, when a hook beats a prompt, and how sessions resume and fork. 27% of the CCAR-F exam.

This is the largest domain in any of the four Claude exams: 27% of CCAR-F, about 16 of the 60 questions. It is also the domain the other four lean on. Tool design (Domain 2) only makes sense once you know how an agent decides to call a tool; context management (Domain 5) is mostly about what a subagent does and does not see.

The vocabulary is small. An agent is Claude running in a loop: it reads a request, decides whether it needs a tool, calls it, reads the result, and decides again, until the job is done. A coordinator is an agent whose tools include starting other agents, called subagents. A hook is a piece of ordinary code that runs at a fixed moment in that loop, before or after a tool call, and can change or block what happens. A session is the saved state of one conversation, which can be resumed or forked.

The exam's questions in this domain come mostly from two of the six scenarios: the customer support agent that must verify a customer before touching a refund, and the multi-agent research system whose coordinator delegates to search, analysis and synthesis subagents. Read both scenarios on the CCAR-F exam guide page before starting; the lessons refer to them by name.

The recurring test is the same throughout: when does a guarantee matter more than a good chance? A prompt instruction gives Claude a good chance of doing the right thing. A hook, a prerequisite gate or a stop_reason check gives a guarantee. Many questions are decided by knowing which of the two the situation demands.

What the exam guide tests

The official CCAR-F guide lists 7 task statements for this domain. Exam questions are written against them, and so are the lessons: each row says where it is covered.

Lessons

1. 1.1 Designing agentic loopsWhy an agent is a loop, who decides what happens in it, why stop_reason is the only signal your code should trust, and the loop mistakes the exam asks about.21 min
2. 1.2 Multi-agent orchestration: the coordinator and its subagentsHow a coordinator splits a research question, routes all messages through itself, briefs subagents that start blank, and loops until coverage is complete.19 min
3. 1.3 Spawning subagents: the Task tool, context passing and forkingHow a coordinator spawns subagents with the Task tool, why "Task" must be in allowedTools, what an AgentDefinition holds, and how findings reach a subagent.21 min
4. 1.4 Workflow enforcement and handoff: guarantees, not good intentionsWhy a prompt saying "verify first" still fails, how a prerequisite gate blocks process_refund until get_customer succeeds, and what a handoff must contain.20 min
5. 1.5 Agent SDK hooks: intercepting tool calls and normalising dataHow Agent SDK hooks enforce rules: a PostToolUse hook normalises mixed MCP results, a PreToolUse hook blocks a $500 refund, and when a hook beats a prompt.22 min
6. 1.6 Task decomposition: fixed pipelines versus adaptive plansWhen to split a job into a fixed prompt chain and when to map first and plan as you go, taught on a 12-file code review and a legacy test job.20 min
7. 1.7 Sessions: resuming, forking and starting freshHow a Claude session is resumed by name, forked from a shared baseline, told which files changed, and when a fresh session with a summary beats resuming.20 min

Practice question

From the CCAR-F bank, tagged to this domain. Every answer option is explained. Nothing is stored, nothing to sign up for.