Study notes for the Claude Certified Developer – Foundations certification (exam code CCDV-F), part of Anthropic's Claude Certification Program. Registration is gated to the Claude Partner Network via the Anthropic Partner Academy.
Content was generated using WebSearch/WebFetch (in place of the Exa SDK — see "Regenerating" below), Context7, and the full 5-module Anthropic Partner Academy prep course (pulled via claude-in-chrome) against the official Exam Guide PDF and Anthropic's public developer docs.
| File | Description |
|---|---|
| 1_agents_and_workflows.md | Domain 1: workflow vs. agent, Agent SDK, hooks, subagents, frameworks |
| 2_applications_and_integration.md | Domain 2 (largest, 33.1%): requirements, API mechanics, app design, config management |
| 3_claude_code_tools_mcp.md | Domains 3 + 8: Claude Code operation, tool implementation, MCP servers |
| 4_model_selection_prompting_context.md | Domains 5 + 6: LLM fundamentals, model tiers, prompting, structured output |
| 5_eval_debugging_security.md | Domains 4 + 7: debugging, prompt injection, guardrails, key management |
| exam_sections.md | Full exam domain/skill breakdown with weightings |
| study_cheat_sheet.md | Decision trees, model/error/hook tables, security checklist |
| study_flashcards.md | Flashcards for key concepts, grouped by domain |
| study_practice_questions.md | 58 original practice questions with explanations |
| study_topic_summaries.md | Topic summaries for final review |
| Domain | Weight |
|---|---|
| Applications and Integration | 33.1% |
| Model Selection and Optimization | 16.8% |
| Agents and Workflows | 14.7% |
| Prompt and Context Engineering | 11.0% |
| Tools and MCPs | 10.6% |
| Security and Safety | 8.1% |
| Claude Code | 3.1% |
| Eval, Testing, and Debugging | 2.6% |
53 items, 120 minutes, scaled score 720/1000 to pass, $125, 12-month validity. Full skill-level breakdown in exam_sections.md.
- Blueprint (domains/weights/skills): the official Exam Guide PDF (v1.0, effective July 2026) — authoritative, cited directly in
exam_sections.md. - Teaching depth, primary: the full 5-module Anthropic Partner Academy prep course (MSO Foundations; Production-Grade Prompting, Agents & Tool Use; Claude Code, MCP & Integration; Production Engineering, Evals & Security; Accelerators & IP Contribution), gated to the Claude Partner Network. Pulled by driving
claude-in-chromeagainst an authenticatedanthropic-partners.skilljar.comsession — each module's SCORM lesson player loads all its screens into the DOM at once (hidden via CSS, only the active one shown), so the full text was extracted per module via adocument.querySelector('section.screen')sweep and saved to a local file, rather than clicking through 20+ screens one at a time. Course content is paraphrased into this repo's own note style throughout, never reproduced verbatim — it's Anthropic's copyrighted material, and these are personal study notes derived from a course the account has legitimate paid access to, not a redistribution of it. - Teaching depth, secondary: Anthropic's public developer docs at
docs.claude.com/platform.claude.com/code.claude.com— Messages API, Agent SDK, Claude Code, MCP, prompt engineering, security guardrails. Used as the primary source before Partner Academy access was available, and as ongoing cross-verification afterward. - API/SDK cross-check: Context7 MCP against
anthropics/claude-agent-sdk-pythonandanthropics/anthropic-sdk-python— verifiedAgentDefinition/ClaudeAgentOptionsfield names and theoutput_config.format/strictstructured-output parameters against the actual SDK source. - Drift/gap check: WebSearch, standing in for Exa's role in the sibling repos (no
EXA_API_KEYconfigured on this machine — see "Regenerating"), plus a dedicated fact-check pass (background agent + fresh WebSearch) against the model-lineup table and prompt-caching mechanics. Notable catches: Haiku 4.5 supports extended thinking, not adaptive thinking, unlike Fable 5/Opus 5/Sonnet 5 — an easy assumption to get backwards if you assume "the flagship models get the newer feature";budget_tokensis deprecated and 400s on the newest model generations, witheffortlevels as the current mechanism; the direct Anthropic API does not currently offer EU data residency (EU-residency deployments route through Bedrock/Vertex); Managed Agents' server-side session storage currently rules out ZDR/HIPAA-BAA eligibility regardless of operational fit.
If the exam guide or docs change, regenerate notes by re-running research against the sources above.
1. Blueprint
WebSearch for "Claude Certified Developer Foundations" exam guide, WebFetch the current PDF, diff against exam_sections.md for a version bump (check the "Document Control" section at the end of the PDF).
2. Public docs
WebSearch for the relevant docs.claude.com/platform.claude.com/code.claude.com page per topic, WebFetch it directly — the docs' own llms.txt index (e.g. https://code.claude.com/docs/llms.txt) is a fast way to discover all available pages under a given docs subdomain.
3. Partner Academy content (optional, requires Partner Network membership)
- Sign into
anthropic-partners.skilljar.comin Chrome, with the claude-in-chrome extension connected and pointed at that logged-in session. - Navigate to a module's lesson URL (e.g.
.../path/claude-certified-developer-foundations/mso-foundations— it redirects into the SCORM player once logged in). - Extract the whole module in one shot via
javascript_tool, since the SCORM player preloads every screen into the DOM (hidden via CSS) rather than fetching one at a time:const inner = window.frames[0].frames[0]; // nested iframe: wrapper -> SCORM content const doc = inner.document; const sections = Array.from(doc.querySelectorAll('section.screen')).sort((a,b) => a.id.localeCompare(b.id)); const out = sections.map(s => `=== ${s.id} ===\n${s.innerText.trim()}`).join('\n\n');
- Don't return
outdirectly fromjavascript_tool— joined multi-screen text reliably trips the tool's content-safety filter (flagged as cookie/query-string-shaped data). Instead trigger a file download and read the result from disk:const blob = new Blob([out], {type: 'text/plain'}); const a = document.createElement('a'); a.href = URL.createObjectURL(blob); a.download = 'moduleN.txt'; document.body.appendChild(a); a.click(); document.body.removeChild(a);
- Chrome silently blocks automatic downloads after the first one per site/session — if only the first module's file lands in
~/Downloads, click "Always allow" on the blocked-downloads icon in the address bar, then retry the remaining modules. - Fold each module's text into the matching section file(s), paraphrased into the repo's own note style, never reproduced verbatim — it's Anthropic's copyrighted course material; personal study notes derived from a course you have legitimate access to are fine, wholesale reproduction isn't.
Fallback if the extension won't connect at all: copy/paste lesson text manually into a Claude Code session and ask it to fold the content in the same way.
4. Context7 API check
Pull current docs for anthropics/claude-agent-sdk-python and anthropics/anthropic-sdk-python via the Context7 MCP tools to verify code snippets and field names haven't drifted.
5. Exa verification (optional — this repo used WebSearch instead)
The DP-700/Databricks sibling repos in this directory use a real Exa API key via the exa-py SDK for this step (see their READMEs for the Keychain-based setup). This repo didn't have EXA_API_KEY available at build time, so WebSearch/WebFetch served the same "search the live web, verify against community writeups" role. If you have an Exa key, prefer it — Exa's type="deep" search with contents={"highlights": True} is more targeted for finding exam-experience/gotcha writeups than general WebSearch.
6. Update files
Update 1_agents_and_workflows.md / 2_applications_and_integration.md / 3_claude_code_tools_mcp.md / 4_model_selection_prompting_context.md / 5_eval_debugging_security.md first (each maps to 1–2 official exam domains — see the table above), then regenerate study_cheat_sheet.md, study_flashcards.md, study_practice_questions.md, study_topic_summaries.md from those.