CCA-F - 4-Week Practice Plan¶
Overview¶
This 4-week study plan is designed for someone with 6+ months of production Claude experience. Adjust the pace based on your background. Each week includes specific tasks, documentation to read, courses to complete, and hands-on exercises.
Time commitment: 2-3 hours per day, 5-6 days per week
Week 1 - API Fundamentals & Prompt Engineering¶
Focus: Domains 3 (Prompt Engineering - 20%) and Domain 5 (Context & Reliability - 15%)
Anthropic Academy Courses¶
- Complete "Claude 101" on Skilljar
- Complete "AI Fluency Framework and Foundations" on Skilljar
- Complete "Building Applications with Claude API" on Skilljar
- Complete "Prompt Engineering with Claude" on Skilljar
Documentation Reading¶
- Read Getting Started Guide
- Read Messages API Reference
- Read Prompt Engineering Overview
- Read System Prompts Guide
- Read Extended Thinking
- Read Prompt Caching
- Read Streaming Guide
- Read Vision Guide
- Read PDF Support
Hands-On Exercises¶
- Set up Anthropic API key and make first API call
- Write 5 different system prompts for different use cases
- Practice chain-of-thought prompting with 3 complex tasks
- Use XML tags to organize a multi-section prompt
- Implement few-shot prompting with 3+ examples
- Extract structured JSON from unstructured text (3 examples)
- Experiment with prompt caching on a repeated prompt
- Test extended thinking on a complex reasoning task
- Implement streaming for a real-time response
Study Notes Review¶
Week 1 Self-Assessment¶
Answer these questions without looking at notes:
- What is the difference between system prompts and user prompts?
- When should you use XML tags in prompts?
- How does prompt caching work and when should you use it?
- What is extended thinking and when is it beneficial?
- How do you handle streaming responses?
- What are the token limits for Claude's context window?
Week 2 - Tool Use & MCP¶
Focus: Domain 4 (Tool Design & MCP Integration - 18%) and Domain 2 (Claude Code - 20%)
Anthropic Academy Courses¶
- Complete "Introduction to Model Context Protocol" on Skilljar
- Complete "Tool Use with Claude" on Skilljar
- Complete "Claude Code Fundamentals" on Skilljar
- Complete "Advanced Claude Code" on Skilljar
Documentation Reading¶
- Read Tool Use Guide
- Read Tool Use Best Practices
- Read MCP Introduction
- Read MCP Core Architecture
- Read MCP Tools
- Read MCP Resources
- Read MCP Transports
- Read Claude Code Overview
- Read Claude Code Memory (CLAUDE.md)
- Read Claude Code Settings
- Read Claude Code Hooks
- Read Claude Code Slash Commands
Hands-On Exercises¶
- Define 3 tools and use them via the Messages API
- Implement parallel tool use in a single request
- Test all tool choice modes (auto, any, specific)
- Handle tool use errors gracefully
- Build a simple MCP server (TypeScript or Python) with 2+ tools
- Connect your MCP server to Claude Code
- Create a CLAUDE.md file for a real project
- Set up a pre-tool hook in Claude Code
- Create a custom slash command
- Configure Claude Code IDE integration (VS Code or JetBrains)
Study Notes Review¶
Week 2 Self-Assessment¶
- What are the three tool choice modes and when do you use each?
- What is the difference between MCP tools, resources, and prompts?
- What transport types does MCP support?
- How does the CLAUDE.md file hierarchy work?
- What are hooks and when would you use them?
- How do you handle errors from tool calls?
Week 3 - Agentic Architecture & Claude Code Advanced¶
Focus: Domain 1 (Agentic Architecture - 27%) - the highest-weighted domain
Anthropic Academy Courses¶
- Complete "Building Agentic Applications" on Skilljar
- Complete "Multi-Agent Systems" on Skilljar
- Complete "Production AI Systems" on Skilljar
- Complete "Responsible AI with Claude" on Skilljar
Documentation Reading¶
- Read Agentic Patterns
- Read Computer Use
- Read Claude Code Best Practices
- Read Rate Limits
- Read API Errors Reference
- Read Anthropic Cookbook - Agentic examples
- Read Token Counting
- Review Claude Code MCP Configuration
Hands-On Exercises¶
- Build an agentic workflow with a tool use loop
- Implement plan-execute-reflect pattern
- Add error handling and retry logic to your agent
- Build a multi-agent system with supervisor pattern
- Implement context summarization for a long-running agent
- Measure and optimize cost for an agentic workflow
- Set up Claude Code for a team project with shared CLAUDE.md
- Implement a production reliability pattern (fallback chain)
- Build an agent that knows when to stop and ask for clarification
Study Notes Review¶
- Review notes/01-agentic-architecture.md
- Re-review all previous notes for cross-domain connections
Week 3 Self-Assessment¶
- When should you use an agent vs a simple prompt chain?
- What is the supervisor pattern and when is it appropriate?
- How do you handle errors in an agentic loop?
- What are the cost implications of extended thinking in agents?
- How do you prevent infinite loops in agentic systems?
- What is the plan-execute-reflect pattern?
Week 4 - Practice Exams & Review¶
Focus: All domains - consolidation and gap filling
Remaining Anthropic Academy Courses¶
- Complete any remaining Skilljar courses not yet finished
- Review course notes and key takeaways from all courses
Practice and Review¶
- Complete all scenarios in scenarios.md
- Time yourself on scenarios (aim for under 2 min per question)
- Review fact-sheet.md end to end
- Re-read strategy.md exam day tactics section
- Review notes/06-exam-tips-prep-strategy.md
Gap Analysis¶
- List topics where you feel least confident
- Re-read documentation for weak areas
- Build one more hands-on example for each weak area
- Write explanations of key concepts in your own words
Domain-Specific Review¶
Domain 1 - Agentic Architecture (27%) - [ ] Can design single and multi-agent architectures - [ ] Know when to use agents vs simple prompts - [ ] Understand production cost and latency tradeoffs - [ ] Can implement error handling and recovery patterns
Domain 2 - Claude Code Configuration (20%) - [ ] Know CLAUDE.md file hierarchy and precedence - [ ] Can configure hooks, slash commands, and settings - [ ] Understand permissions model - [ ] Can set up MCP servers in Claude Code
Domain 3 - Prompt Engineering & Structured Output (20%) - [ ] Can write effective system prompts - [ ] Know when to use XML tags, chain-of-thought, few-shot - [ ] Can extract structured JSON from unstructured text - [ ] Understand prompt caching strategies
Domain 4 - Tool Design & MCP Integration (18%) - [ ] Can design tools with proper schemas and descriptions - [ ] Understand MCP architecture and transport types - [ ] Can build MCP servers with tools and resources - [ ] Know tool choice modes and parallel tool use
Domain 5 - Context & Reliability (15%) - [ ] Understand context window management strategies - [ ] Can implement reliability patterns (retries, fallbacks) - [ ] Know extended thinking capabilities and tradeoffs - [ ] Can handle rate limits and API errors
Exam Logistics¶
- Schedule exam through ProctorFree
- Test ProctorFree setup and environment
- Prepare quiet, clean room for proctoring
- Have government ID ready
- Get a good night's sleep before exam day
Daily Study Template¶
Use this template for each study session:
Date: ___________
Time spent: ___________
Topics covered:
- [ ] ___________
- [ ] ___________
- [ ] ___________
Key takeaways:
1. ___________
2. ___________
3. ___________
Questions to research:
1. ___________
2. ___________
Confidence level (1-5): ___
Progress Summary¶
| Week | Focus | Status |
|---|---|---|
| Week 1 | API Fundamentals & Prompt Engineering | [ ] Not Started / [ ] In Progress / [ ] Complete |
| Week 2 | Tool Use & MCP | [ ] Not Started / [ ] In Progress / [ ] Complete |
| Week 3 | Agentic Architecture | [ ] Not Started / [ ] In Progress / [ ] Complete |
| Week 4 | Practice & Review | [ ] Not Started / [ ] In Progress / [ ] Complete |
Exam Date: __ Result: __