Claude Certification (CCAR-F)

Thirty lessons for the Claude Certified Architect – Foundations (CCAR-F) exam, crawled from the curriculum and lookup sections of claudecertificationguide.com/learn.

Start here

CCAR-F Exam Overview60 items in 120 minutes, a 720 pass mark, the five domain weights, retake policy and registration route.

Domain 1 — Agentic Architecture & Orchestration (27%)

1.1 Agentic LoopsDesign and implement agentic loops for autonomous task execution using the Messages API, stop_reason field, and tool result handling.

1.2 Multi-Agent OrchestrationOrchestrate multi-agent systems with coordinator-subagent patterns, hub-and-spoke architecture, and proper isolation principles.

1.3 Subagent Invocation and Context PassingConfigure subagent invocation using the Task tool (renamed Agent in current Claude Code), explicit context passing with structured metadata, parallel spawning, and fork_session.

1.4 Workflow Enforcement and HandoffImplement multi-step workflows with programmatic enforcement, prerequisite gates, and structured handoff protocols for human escalation.

1.5 Agent SDK HooksApply Agent SDK hooks for data normalisation and deterministic policy enforcement using the PostToolUse and PreToolUse hook events.

1.6 Task Decomposition StrategiesDesign task decomposition strategies using fixed sequential pipelines and dynamic adaptive decomposition, and solve the attention dilution problem.

1.7 Session State and ResumptionManage session state using named session resumption, fork_session for divergent exploration, and fresh starts with summary injection to avoid stale context.

Domain 2 — Tool Design & MCP Integration (18%)

2.1 Tool Interface DesignDesigning effective tool interfaces with clear descriptions and boundaries for reliable LLM tool selection

2.2 Structured Error ResponsesImplementing structured error responses for MCP tools with proper categorisation and recovery metadata

2.3 Tool Distribution & Tool ChoiceDistributing tools across agents and configuring tool_choice for reliable tool selection

2.4 MCP Server IntegrationIntegrating MCP servers into Claude Code and agent workflows with proper scoping and configuration

2.5 Built-in ToolsSelecting and applying built-in tools (Read, Write, Edit, Bash, Grep, Glob) effectively for codebase operations

Domain 3 — Claude Code Configuration & Workflows (20%)

3.1 CLAUDE.md Hierarchy, Scoping, and Modular OrganisationMaster the three-level CLAUDE.md configuration hierarchy, understand scoping rules, and learn modular organisation patterns with @path imports and .claude/rules/

3.2 Custom Slash Commands and SkillsLearn the unified Skills system for custom commands, configure SKILL.md frontmatter, and understand when to use skills vs CLAUDE.md

3.3 Path-Specific Rules for Conditional Convention LoadingMaster path-specific rules in .claude/rules/ with YAML frontmatter glob patterns for conditional convention loading, token efficiency, and cross-directory coverage

3.4 Plan Mode vs Direct ExecutionLearn the decision framework for choosing plan mode or direct execution, understand the Explore subagent, and master the hybrid plan-then-execute pattern

3.5 Iterative Refinement TechniquesMaster the technique hierarchy for iterative refinement: concrete examples over prose, test-driven iteration, the interview pattern, and batch vs sequential feedback

3.6 CI/CD IntegrationMaster running Claude Code in CI/CD pipelines: the -p flag for non-interactive mode, structured JSON output, session isolation for reviews, and incremental review context

Domain 4 — Prompt Engineering & Structured Output (20%)

4.1 System Prompts with Explicit CriteriaDesign prompts with explicit criteria to improve precision and reduce false positives

4.2 Few-Shot PromptingApply few-shot prompting to improve output consistency and quality

4.3 Structured Output with Tool UseEnforce structured output using tool use and JSON schemas

4.4 Validation, Retry, and Feedback LoopsImplement validation, retry, and feedback loops for extraction quality

4.5 Batch Processing StrategiesDesign efficient batch processing strategies

4.6 Multi-Instance and Multi-Pass ReviewDesign multi-instance and multi-pass review architectures

Domain 5 — Context Management & Reliability (15%)

5.1 Context Window ManagementManage conversation context to preserve critical information across long interactions, including the progressive summarisation trap, lost-in-the-middle effect, and tool result trimming.

5.2 Escalation & Ambiguity ResolutionDesign effective escalation and ambiguity resolution patterns, including valid escalation triggers, unreliable triggers, frustration nuance, and ambiguous customer matching.

5.3 Error Propagation in Multi-Agent SystemsImplement error propagation strategies across multi-agent systems, including structured error context, the distinction between access failures and valid empty results, and coverage annotations.

5.4 Codebase Exploration & Context DegradationManage context effectively in large codebase exploration, including context degradation mitigation, scratchpad files, subagent delegation, and crash recovery via structured state manifests.

5.5 Human Review & Confidence CalibrationDesign human review workflows and confidence calibration, including the aggregate metrics trap, stratified random sampling, field-level confidence scores, and reviewer capacity prioritisation.

5.6 Information Provenance & Multi-Source SynthesisPreserve information provenance and handle uncertainty in multi-source synthesis, including structured claim-source mappings, conflict handling, temporal awareness, and content-appropriate rendering.

Look up

Quick ReferenceCheat sheets for all five domains: decision-rule tables, numbers to memorise and structural templates.

GlossaryEvery CCAR-F term defined and grouped by domain, each with the exam context behind it.

Common Exam TrapsAll 117 traps flagged across the 30 lessons, grouped by domain, with the reason each distractor fails.

Content crawled from the publicly available material at claudecertificationguide.com (an independent site, not affiliated with Anthropic), cross-checked against docs.anthropic.com. For personal study only.