CCDV-F Study Lab
A study site for Anthropic's Claude Certified Developer exam, with practice questions checked against the official docs and tests scored by exam objective.
- result
- Published a free study site with 313 doc-backed practice questions and weekly source checks. I used it while preparing for the exam, which I passed with 896/1000.
- my contribution
- Built the study site, question bank, objective-scored tests, and local agent playground.
- 2026–Present
- web
how it works
I built CCDV-F Study Lab to prepare for Anthropic's Claude Certified Developer – Foundations exam, which I passed with 896/1000. Its bank holds 313 practice questions, each written from a page of Anthropic's docs and stored with the exact quote that backs the answer. A weekly GitHub Actions job fetches every source page and fails if a quote is gone, so questions don't go stale as the docs change.
The public site is a static Next.js export on GitHub Pages: knowledge checks, a 53-question mock exam weighted like the real one, custom tests scored by the exam's 25 objectives, a mistakes deck and a study plan, with progress kept in the browser. Run locally, the same codebase is a playground where Claude Code works on sample projects through the Claude Agent SDK, with 27 challenges checked automatically.
engineering details
- Wrote a 313-question bank where every answer carries a quote from the official docs, re-checked weekly in CI.
- Built mock exams and custom tests that spread questions across the exam's 25 objectives and score each one, like the real score report.
- Published the study track as a static export with every API route left out of the build, redeployed to GitHub Pages on each push.
- Added optional GitHub sign-in that syncs progress through Supabase, with row-level security limiting each user to their own row.
- Covered the app with 720 Jest tests, including a local practice Claude API so the coding challenges run without an API key.
in the product
Objective-Scored Practice Tests
Pick a length and the objectives to cover; results show percent correct per objective, with weak ones flagged.
Doc-Checked Question Bank
Every question links to its source page and stores the quote behind its answer; a weekly job checks each quote is still there.
Agent Playground
Locally, Claude Code edits a sample project through the Agent SDK while each tool call, permission prompt and message shows up live.