Reference · completeness audit
Coverage Matrix
Every one of the 28 task statements mapped to where it's taught, after a line-by-line review of each "Knowledge of" and "Skills in" bullet in the Exam Guide. ★ marks content added or enriched in the completeness pass.
Audit resultAll 28 task statements and all appendix technologies are covered. The completeness pass added a scenario playbook, an API/SDK essentials reference, a hands-on labs lesson, and enriched AgentDefinition, CI test-generation, and glossary entries (max_tokens, semantic-validation fields).
Domain 1 — Agentic Architecture & Orchestration (27%)
| Task | Covered in | Status |
|---|
| 1.1 Agentic loops | L1 (0001) | ✓ |
| 1.2 Coordinator–subagent patterns | L2 (0002) | ✓ |
| 1.3 Subagent invocation/context/spawning | L2 (0002) — ★ AgentDefinition expanded | ✓ |
| 1.4 Multi-step workflows: enforcement/handoff | L3 (0003) | ✓ |
| 1.5 Agent SDK hooks | L3 (0003) | ✓ |
| 1.6 Task decomposition strategies | L4 (0004) | ✓ |
| 1.7 Session state/resumption/forking | L4 (0004) | ✓ |
Domain 2 — Tool Design & MCP Integration (18%)
| Task | Covered in | Status |
|---|
| 2.1 Tool interface design | L1 (0006) | ✓ |
| 2.2 Structured MCP errors | L2 (0007) | ✓ |
| 2.3 Tool distribution & tool_choice | L3 (0008) | ✓ |
| 2.4 MCP servers into Claude Code | L4 (0009) | ✓ |
| 2.5 Built-in tools | L4 (0009) | ✓ |
Domain 3 — Claude Code Configuration & Workflows (20%)
| Task | Covered in | Status |
|---|
| 3.1 CLAUDE.md hierarchy/scoping/modularity | L1 (0011) | ✓ |
| 3.2 Custom slash commands & skills | L2 (0012) | ✓ |
| 3.3 Path-specific rules | L1 (0011) | ✓ |
| 3.4 Plan mode vs direct execution | L3 (0013) | ✓ |
| 3.5 Iterative refinement | L4 (0014) | ✓ |
| 3.6 CI/CD integration | L4 (0014) — ★ test-generation section added | ✓ |
Domain 4 — Prompt Engineering & Structured Output (20%)
| Task | Covered in | Status |
|---|
| 4.1 Explicit criteria / false positives | L1 (0016) | ✓ |
| 4.2 Few-shot prompting | L1 (0016) | ✓ |
| 4.3 Structured output via tool_use + schemas | L2 (0017) | ✓ |
| 4.4 Validation/retry/feedback loops | L2 (0017) — ★ detected_pattern in glossary | ✓ |
| 4.5 Batch processing | L3 (0018) | ✓ |
| 4.6 Multi-instance / multi-pass review | L3 (0018) | ✓ |
Domain 5 — Context Management & Reliability (15%)
| Task | Covered in | Status |
|---|
| 5.1 Preserve context across long interactions | L1 (0020) | ✓ |
| 5.2 Escalation & ambiguity resolution | L2 (0021) | ✓ |
| 5.3 Error propagation across agents | L2 (0021) | ✓ |
| 5.4 Context in large codebase exploration | L1 (0020) | ✓ |
| 5.5 Human review & confidence calibration | L3 (0022) | ✓ |
| 5.6 Provenance & uncertainty in synthesis | L3 (0022) | ✓ |
Appendix technologies & the 6 scenarios
| Item | Covered in | Status |
|---|
Appendix API/SDK primitives (stop_reason values, max_tokens, system prompts, tool_choice, Pydantic, Batches, JSON Schema) | ★ API & SDK Essentials + across lessons | ★ new |
| All 6 exam scenarios (setup, tools, domains, traps) | ★ Scenario Playbook | ★ new |
| Preparation exercises 1–4 | ★ Hands-On Labs (0025) | ★ new |
| Out-of-scope topics | MISSION.md + API/SDK Essentials | ✓ |
Honest caveats
Content is grounded in the official Exam Guide (high confidence on concepts, field names, CLI flags). Live docs.claude.com URLs cited as "primary source" reading have not each been re-verified to resolve today — ask your teacher to confirm any specific link. The mock-exam question count, timer, and scaled-score formula are sized to a typical proctored cert, since Anthropic does not publish exact count/time/scaling.