Anthropic Claude Certified Architect – Professional (CCAR-P) - December 7 onwards Lessons
39 lessons, one per exam-guide subdomain, in the order the guide teaches them. Every claim is cited to the official documentation.
Domain 1: Solution Design and Architecture
6 lessons · 18% of the exam
1.1Delegating process steps between Claude and humans
Subdomain 1.1: Task Delegation Between Claude and Humans
15 min read
1.2End-to-end architecture evaluation: finding the component that fails first
Subdomain 1.2: End-to-End Architecture Evaluation
17 min read
1.3Choosing between workflow, agent, augmented-LLM and Skills patterns
Subdomain 1.3: Architectural Pattern Selection
19 min read
1.4Evaluating orchestrator and subagent decompositions
Subdomain 1.4: Orchestrator and Subagent Decomposition
18 min read
1.5Value-based architectural decisions for Claude solutions
Subdomain 1.5: Value-Based Architectural Decisions
16 min read
1.6Choosing a Claude deployment surface and agent runtime
Subdomain 1.6: Deployment Surface and Agent Runtime Selection
18 min read
Domain 2: Claude Models, Prompting, and Context Engineering
6 lessons · 13% of the exam
2.1Choosing a Claude model tier for a workload's binding constraints
Subdomain 2.1: Model Tier Selection
16 min read
2.2Diagnosing and revising underspecified system prompts
Subdomain 2.2: System Prompt Diagnosis and Revision
16 min read
2.3Prompting technique selection: few-shot, thinking, structured output and chaining
Subdomain 2.3: Prompting Technique Selection
17 min read
2.4Token budgets: auditing context composition and cache behaviour
Subdomain 2.4: Context Composition and Token Economics
17 min read
2.5Prompt cache breakpoints and stable-prefix packaging
Subdomain 2.5: Cache Breakpoints and Reusable Packaging
19 min read
2.6Reasoning effort and context lifecycle for long-running agents
Subdomain 2.6: Reasoning Effort and Context Lifecycle
20 min read
Domain 3: Integration
8 lessons · 21% of the exam
3.1Least-privilege tool configuration for agents and subagents
Subdomain 3.1: Least-Privilege Tool Configuration
16 min read
3.2Agent credential exposure and privilege escalation
Subdomain 3.2: Authentication and Authorization Posture
16 min read
3.3Relieving the binding constraint: accuracy, latency or throughput
Subdomain 3.3: Relieving the Binding Performance Constraint
16 min read
3.4Instrumentation gaps in agentic systems: finding the signal that would have fired
Subdomain 3.4: Instrumentation and Monitoring Gaps
16 min read
3.5Retrieval architecture selection: full context, agentic search or RAG
Subdomain 3.5: Retrieval Architecture Selection
19 min read
3.6Choosing between direct tool calls, MCP and agent handoff
Subdomain 3.6: Connection Mechanism Selection
17 min read
3.7Progressive discovery vs. loading everything up front
Subdomain 3.7: Progressive Discovery vs. Monolithic Context Loading
18 min read
3.8Production reliability controls: rate limits, retries, idempotency and fallback
Subdomain 3.8: Production Reliability Controls
18 min read
Domain 4: Evaluation, Testing, and Optimization
7 lessons · 17% of the exam
4.1Evaluation coverage gaps: finding the metric that would have caught it
Subdomain 4.1: Evaluation Coverage Gaps
17 min read
4.2Deriving evaluations from business processes
Subdomain 4.2: Deriving Evaluations from Business Processes
16 min read
4.3LLM-as-judge bias: diagnosing and correcting position, verbosity, self-preference and drift
Subdomain 4.3: LLM-as-Judge Design and Bias
16 min read
4.4Comparing two Claude configurations: trials, cases and run-to-run noise
Subdomain 4.4: Experimental Design for Configuration Comparison
16 min read
4.5Root-cause isolation from Claude logs and traces
Subdomain 4.5: Root-Cause Isolation from Logs and Traces
16 min read
4.6Cost and latency optimization from usage telemetry
Subdomain 4.6: Cost and Latency Optimization
18 min read
4.7Model version migration: deprecation timelines, regressions and drift
Subdomain 4.7: Model Version Migration
17 min read
Domain 5: Governance, Safety, and Risk Management
6 lessons · 16% of the exam
5.1Guardrail placement: closing a misuse path without breaking the legitimate use
Subdomain 5.1: Guardrail Placement
17 min read
5.2Prompt injection surface and controls for Claude agents
Subdomain 5.2: Prompt Injection Surface and Controls
18 min read
5.3Risk assessment for Claude deployments: ranking exposure
Subdomain 5.3: Risk Assessment and Ranking
17 min read
5.4Human-in-the-loop approval gates for agent tool calls
Subdomain 5.4: Human-in-the-Loop Checkpoints
17 min read
5.5Compliance controls for Claude: residency, ZDR, PHI and audit evidence
Subdomain 5.5: Architectural Controls for Compliance Obligations
21 min read
5.6Bias source identification and mitigation
Subdomain 5.6: Bias Source Identification and Mitigation
16 min read
Domain 6: Stakeholder Communication and Lifecycle Management
3 lessons · 7% of the exam
6.1Unsupportable commitments in Claude requirements and SLAs
Subdomain 6.1: Unsupportable Commitments in Requirements and SLAs
17 min read
6.2Architecture decision records for Claude systems: trade-offs and residual risk
Subdomain 6.2: Architecture Decision Records
16 min read
6.3Operational handoff completeness for Claude systems
Subdomain 6.3: Operational Handoff Completeness
18 min read
Domain 7: Developer Productivity and Operational Enablement
3 lessons · 8% of the exam
7.1Packaging and distributing Claude Code workflows with plugins and managed settings
Subdomain 7.1: Packaging and Distributing Claude Code Workflows
22 min read
7.2Configuring Claude Code and the Agent SDK: permission modes, hooks, subagents and CI
Subdomain 7.2: Agent SDK and Claude Code Configuration
16 min read
7.3Human ownership of AI-generated code: review gates, CI agents and secrets
Subdomain 7.3: Accountability for AI-Generated Code
17 min read