Developer overview
Developer Guide·25 / 26·5 min read
Coverage Map
Summary: Every skill on the published blueprint, its weight, and the chapter that owns it. Use this to find the chapter behind a weakness rather than reading in order. The weights are the blueprint's own and they are published per skill rather than per domain, which is finer than most exams give you. Spend an evening looking at them before you decide where the rest of your evenings go.
Domain 1: Agents and Workflows
| Skill | Weight | Owned by |
|---|---|---|
| 1.1 Agent Architecture | 4.5% | Chapter 12, Known Steps Take a Chain; Unknown Ones an Agent |
| 1.2 Agent Construction with Claude | 5.3% | Chapter 13, The Loop, the Harness, and What You Stop Owning |
| 1.3 Agent Patterns and Frameworks | 4.9% | Chapter 12, Known Steps Take a Chain; Unknown Ones an Agent |
Domain 2: Applications and Integration
| Skill | Weight | Owned by |
|---|---|---|
| 2.1 Understanding Requirements | 3.4% | Chapter 15, The Constraints Nobody States Decide the Design |
| 2.2 Systems Life Cycle | 2.8% | Chapter 23, Deploy Is a Phase Boundary, Not the End |
| 2.3 Claude API Mechanics | 6.8% | Chapter 3, Messages, Streaming, Batches: Who Is Waiting Decides |
| 2.4 Software Engineering Foundations | 7.4% | Chapter 17, REST, Versioning, Review: the Exam Beyond Claude |
| 2.5 Claude Application Design | 8.6% | Chapter 16, Boundaries, Sessions, and Where Instructions Land |
| 2.6 Configuration Management | 4.1% | Chapter 18, An Unpinned Version Is an Untracked Change |
Domain 3: Claude Code
| Skill | Weight | Owned by |
|---|---|---|
| 3.1 Claude Code Operation | 3.1% | Chapter 14, Permission Modes and Where the Human Gate Goes |
Domain 4: Eval, Testing, and Debugging
| Skill | Weight | Owned by |
|---|---|---|
| 4.1 Debugging and Error Handling | 2.6% | Chapter 19, A Test Says It Broke; a Trace Says Where |
Domain 5: Model Selection and Optimization
| Skill | Weight | Owned by |
|---|---|---|
| 5.1 LLM Fundamentals | 5.2% | Chapter 1, What This Exam Tests That an Interview Does Not |
| 5.2 Technical Fundamentals | 6.1% | Chapter 2, Four Parts to Every Call, and Three Places It Breaks |
| 5.3 Model Selection and Tradeoffs | 2.7% | Chapter 4, A Model Choice Is Three Numbers You Were Not Given |
| 5.4 Cost and Token Management | 2.8% | Chapter 8, Cost Follows What You Resend, Not What You Ask |
Domain 6: Prompt and Context Engineering
| Skill | Weight | Owned by |
|---|---|---|
| 6.1 Context Engineering | 3.8% | Chapter 7, The Window Is a Budget, and Everything Competes |
| 6.2 Prompt Engineering | 4.6% | Chapter 5, Instruction Placement Is a Trust Decision |
| 6.3 Output Handling | 2.6% | Chapter 6, Schema-Valid Is Not the Same as Correct |
Domain 7: Security and Safety
| Skill | Weight | Owned by |
|---|---|---|
| 7.1 AI Application Security | 3.2% | Chapter 20, The Attack Arrives as Content You Asked For |
| 7.2 Guardrails and Safe Deployment | 2.3% | Chapter 21, A Rule in a Prompt Is Not Enforcement |
| 7.3 Claude Hooks | 1.0% | Chapter 21, A Rule in a Prompt Is Not Enforcement |
| 7.4 Identity, Secrets, and Key Management | 1.6% | Chapter 22, One Credential: What It Reaches, and for How Long |
Domain 8: Tools and MCPs
| Skill | Weight | Owned by |
|---|---|---|
| 8.1 Tool Implementation | 4.4% | Chapter 9, The Description Is the Input, Not the Documentation |
| 8.2 MCP Server Development | 2.1% | Chapter 10, One Server, Many Clients, Transport by Locality |
| 8.3 Agentic Customization | 4.1% | Chapter 11, Four Mechanisms, Decided by What Must Execute |
Where the Weight Is
Reading order follows the order you meet the decisions, not the blueprint, so a domain is spread across the book rather than gathered into one place. If you are rationing time, these are the numbers that should drive it. Domain 2 is worth thirteen of Domain 4.
| Domain | Weight | Chapters covering it |
|---|---|---|
| 1 Agents and Workflows | 14.7% | 12, 13 |
| 2 Applications and Integration | 33.1% | 3, 15, 16, 17, 18, 23 |
| 3 Claude Code | 3.1% | 14 |
| 4 Eval, Testing, and Debugging | 2.6% | 19 |
| 5 Model Selection and Optimization | 16.8% | 1, 2, 4, 8 |
| 6 Prompt and Context Engineering | 11.0% | 5, 6, 7 |
| 7 Security and Safety | 8.1% | 20, 21, 22 |
| 8 Tools and MCPs | 10.6% | 9, 10, 11 |
Every Chapter, in Order, With What It Owns
The other direction, for when you know which chapter you are in and want to know what it is answerable for.
| Ch | Title | Skills | Exam weight |
|---|---|---|---|
| 1 | What This Exam Tests That an Interview Does Not | 5.1 | 5.2% |
| 2 | Four Parts to Every Call, and Three Places It Breaks | 5.2 | 6.1% |
| 3 | Messages, Streaming, Batches: Who Is Waiting Decides | 2.3 | 6.8% |
| 4 | A Model Choice Is Three Numbers You Were Not Given | 5.3 | 2.7% |
| 5 | Instruction Placement Is a Trust Decision | 6.2 | 4.6% |
| 6 | Schema-Valid Is Not the Same as Correct | 6.3 | 2.6% |
| 7 | The Window Is a Budget, and Everything Competes | 6.1 | 3.8% |
| 8 | Cost Follows What You Resend, Not What You Ask | 5.4 | 2.8% |
| 9 | The Description Is the Input, Not the Documentation | 8.1 | 4.4% |
| 10 | One Server, Many Clients, Transport by Locality | 8.2 | 2.1% |
| 11 | Four Mechanisms, Decided by What Must Execute | 8.3 | 4.1% |
| 12 | Known Steps Take a Chain; Unknown Ones an Agent | 1.1, 1.3 | 9.4% |
| 13 | The Loop, the Harness, and What You Stop Owning | 1.2 | 5.3% |
| 14 | Permission Modes and Where the Human Gate Goes | 3.1 | 3.1% |
| 15 | The Constraints Nobody States Decide the Design | 2.1 | 3.4% |
| 16 | Boundaries, Sessions, and Where Instructions Land | 2.5 | 8.6% |
| 17 | REST, Versioning, Review: the Exam Beyond Claude | 2.4 | 7.4% |
| 18 | An Unpinned Version Is an Untracked Change | 2.6 | 4.1% |
| 19 | A Test Says It Broke; a Trace Says Where | 4.1 | 2.6% |
| 20 | The Attack Arrives as Content You Asked For | 7.1 | 3.2% |
| 21 | A Rule in a Prompt Is Not Enforcement | 7.2, 7.3 | 3.3% |
| 22 | One Credential: What It Reaches, and for How Long | 7.4 | 1.6% |
| 23 | Deploy Is a Phase Boundary, Not the End | 2.2 | 2.8% |
Total skills: 25, every one owned by exactly one chapter. The weights sum to 100.0%.
For detailed info, read the official docs
- Agent Skillsplatform.claude.com
- Configure permissionscode.claude.com
- Agent SDK overviewcode.claude.com