Been prepping for the Claude Certified Architect exam and couldn't find any realistic practice questions anywhere, so I spent the last few weeks building my own.
There's a lesson plan and you can use the 10-question speed tier as many times as you want (and it will shuffle the questions). There is a paid tier for those who want to practice the full 2-hr exam with all the questions.
Here are 3 free ones to give you a feel for the difficulty level:
Q1 — Agentic Architecture
A coordinator agent spawns three subagents in parallel. One fails mid-task. What is the correct behavior?
A) Terminate all subagents and restart from scratch
B) Log the failure silently and return partial results
C) Surface the error to the coordinator to handle recovery
D) Retry the failed subagent indefinitely
✅ Answer: C — subagents don't self-recover. The coordinator owns all error handling.
Q2 — Claude Code
You want Claude Code to run your test suite automatically after every file edit. Where do you configure this?
A) A shell alias in your terminal
B) PostToolUse hooks in your Claude Code config
C) A custom slash command in CLAUDE.md
D) .claude/settings.json under postEditHook
✅ Answer: B — PostToolUse hooks are designed exactly for this.
Q3 — Context Management
Your agent processes a 200-page document and produces inconsistent outputs. Expanding the context window doesn't help. Why?
A) The model needs more few-shot examples
B) The system prompt is conflicting with the document
C) Attention dilution — quality degrades as context grows, regardless of window size
D) The document format isn't supported
✅ Answer: C — more context ≠ better attention. Split into focused per-section passes instead.
I built 300 questions total across all 5 domains with timed sessions and a full 60-question mock exam. The sprint above is free with no signup.
Search "Claude Certified Architects" to find it, or drop a comment and I'll DM you the link.
Happy to answer questions about any of the exam domains — good luck to everyone prepping 🙌
Top comments (0)