README.md

Claude Certified Architect — Foundations (CCA-F)

2 min read index source

Claude Certified Architect — Foundations (CCA-F)

Study workspace for the Claude Certified Architect – Foundations exam. Each subfolder is one exam domain (or support material); each has a README.md study guide listing the concepts to know and the common pitfalls the exam tests.

The access-request / enrollment page: https://anthropic.skilljar.com/claude-certified-architect-foundations-access-request

Exam format

Item Detail
Style Multiple choice — one correct answer, three distractors
Structure Scenario-based: questions drawn from a pool of 6 real-world scenario contexts
Duration ~120 minutes
Passing score 720 on a scaled 100–1000 score
Focus Architectural decisions, not parameter trivia — when and why, not just what

The exam does not test API minutiae. It tests judgment: when to delegate to a subagent vs. do the work directly, when to escalate to a human, when a resource beats a tool, when to enforce a rule in code vs. a prompt.

The 5 domains and weightings

# Domain Weight Folder
1 Agentic Architecture & Orchestration 27% 01_agentic_architecture_orchestration/
2 Tool Design & MCP Integration 18% 02_tool_design_and_mcp/
3 Claude Code Configuration & Workflows 20% 03_claude_code_config_workflows/
4 Prompt Engineering & Structured Output 20% 04_prompt_engineering_structured_output/
5 Context Management & Reliability 15% 05_context_management_reliability/

Plus:

  • 06_scenarios/ — the 6 scenario contexts the exam questions are framed around.
  • 07_resources/ — official study modules, reference links, a quick-reference cheat sheet.

Cross-cutting themes (these decide most questions)

  1. Responsibility assignment — the model interprets language and picks among options; application code enforces permissions, compliance, state, retries, idempotency, validation, audit. Tool/schema design shapes behavior. When a question asks “where should X live,” the answer is almost always code/tool enforcement, not a prompt instruction.
  2. Schema-backed output beats free-form text — tool use / structured outputs over “please return JSON.”
  3. Structured error results beat exceptions — isError: true results with categories and retry flags guide the agent; thrown exceptions don’t.
  4. Explicit enforcement beats prompt instructions — code, tool implementation, and hooks enforce; IMPORTANT/NEVER in a prompt does not.
  5. Context is a decision — include what the next step needs, compress what it doesn’t. Capacity ≠ attention.
  6. Provenance and uncertainty are mandatory for high-stakes work — claim + source ID + location + date + confidence.
  7. Staged vs. monolithic — decompose when the next step depends on findings; combine when steps are mechanical.

How to use this folder

  1. Read this README, then 07_resources/README.md for the study plan and official modules.
  2. Work the 5 domain folders in weight order (Domain 1 first — it’s 27%).
  3. For each domain README, study the Key concepts and drill the Common pitfalls — the pitfalls are written the way distractors are written.
  4. Finish with 06_scenarios/: re-read every domain through the lens of each scenario.
  5. Build the hands-on artifacts listed in 07_resources/README.md — the exam rewards people who’ve actually built these.