Claude Certified Developer - Foundations (CCDV-F) — Simulado e guia de estudo

Choose a workflow or an agent, shape manager and subagent hierarchies, and build agents with the Claude Agent SDK, custom loops, hosted or self-hosted deployment, and hooks. Covers tool-use loops, memory, context-window management, and frameworks used for multi-step work.

Heaviest domain. Derive functional and infrastructure requirements, apply the systems life cycle, and use the Claude API: messages, tools, streaming, vision, thinking, caching, batch versus realtime, and third-party invocation. Includes REST, JSON, async code, review, refactoring, instruction boundaries across Claude Code, Desktop, claude.ai, API, and SDKs, plus CLAUDE.md, settings.json, and model pinning.

Operate Claude Code: Rules, Skills, Commands, Agents, Agent Memory, session and headless modes, slash commands, the CLAUDE.md hierarchy, repository initialization, and settings.json. This domain is a small share of the exam.

Identify error types, pick a recovery strategy, read traces to find failure modes, and separate an integration-layer fault from a model-output fault.

Tokens, context windows, sampling, non-determinism, fast mode, extended and adaptive thinking, and zero-shot through multi-shot prompting. Compare Opus, Sonnet, and Haiku on quality, latency, and cost, including behavior changes across releases. Budget tokens and use prompt caching.

Keep the context window from drifting or bloating, prune tool output, compact history, and isolate context with subagents. Place instructions, few-shot examples, and output constraints correctly. Validate structured output and treat confident text as untrusted until checked.

Mitigate prompt injection and jailbreaks, separate untrusted input from instructions, prevent data leakage, handle PII, and apply least privilege. Use guardrails and hooks to block destructive actions. Manage secrets, API keys, and access approval.

Define tool schemas and function calls, write tool descriptions, handle tool errors, and choose client-side or server-side tools and approval patterns. Build an MCP server with resources, tools, and prompts over stdio or sockets. Choose among built-in tools, custom tools, Skills, and MCP.