12-Week Study Plan for Exam Success

Read exam guide domains 1-5 and understand the 6 scenarios

Study agentic loop lifecycle: stop_reason ('tool_use' vs 'end_turn')

Build a minimal agentic loop with the Agent SDK

Study anti-patterns: parsing NL for loop termination, arbitrary iteration caps

Practice Test 1 (Agentic Loops — 10 questions)

Review wrong answers and re-read task statement 1.1

Rest / light review

Study hub-and-spoke architecture, coordinator role, subagent context isolation

Study Task tool for subagent spawning, allowedTools must include 'Task'

Build a coordinator + 2 subagents with explicit context passing

Study parallel subagent execution, fork_session for branched exploration

Study task decomposition pitfalls (overly narrow = coverage gaps)

Practice Test 2 (Multi-Agent Systems — 10 questions)

Rest / review

Study PostToolUse hooks for data normalization and tool call interception

Study programmatic enforcement vs prompt-based guidance (deterministic vs probabilistic)

Build a hook that blocks refunds above $500 and redirects to escalation

Study session management: --resume, fork_session, named sessions, stale context

Study task decomposition: prompt chaining vs dynamic adaptive decomposition

Practice Test 3 (Hooks, Workflows & Sessions — 10 questions)

Rest / review

Study tool description best practices: input formats, examples, edge cases

Study structured error responses: isError, errorCategory, isRetryable

Study tool distribution: 4-5 tools per agent max, scoped tool access

Study MCP server config: .mcp.json (project) vs ~/.claude.json (user)

Study built-in tools: Read, Write, Edit, Bash, Grep, Glob — when to use each

Practice Test 4 (Tool Design & MCP — 10 questions)

Rest / review

Study CLAUDE.md hierarchy: user, project, directory levels

Study @import syntax, .claude/rules/ directory for topic-specific rules

Study custom slash commands (.claude/commands/) vs skills (.claude/skills/)

Study SKILL.md frontmatter: context: fork, allowed-tools, argument-hint

Study path-specific rules: YAML frontmatter with paths glob patterns

Practice Test 5 (Claude Code Config — 10 questions)

Rest / review

Study plan mode vs direct execution decision criteria

Study iterative refinement: concrete examples, TDD iteration, interview pattern

Study CI/CD: -p flag, --output-format json, --json-schema

Study session context isolation in CI (generator vs reviewer)

Study batch processing: Message Batches API, 50% savings, 24h window

Practice Test 6 (Plan Mode & CI/CD — 10 questions)

Rest / review

Study explicit criteria over vague instructions, false positive impact

Study few-shot prompting: 2-4 examples for ambiguous cases

Study tool_use with JSON schemas: guaranteed schema compliance vs semantic errors

Study tool_choice: 'auto' vs 'any' vs forced specific tool

Study schema design: required vs optional, enums with 'other' + detail

Practice Test 7 (Prompt Engineering — 10 questions)

Rest / review

Study validation-retry loops: append specific errors to prompt

Study detected_pattern fields for tracking dismissal patterns

Study batch processing strategy: synchronous for blocking, batch for latency-tolerant

Study self-review limitations: same session retains reasoning context

Study multi-pass review: per-file local analysis + cross-file integration pass

Practice Test 8 (Validation & Multi-Pass — 10 questions)

Rest / review

Study progressive summarization risks, 'lost in the middle' effect

Study 'case facts' blocks, trimming verbose tool outputs, position-aware ordering

Study escalation patterns: customer demands, policy gaps, sentiment ≠ complexity

Study error propagation: structured context vs generic errors

Study local recovery before coordinator escalation, partial results reporting

Practice Test 9 (Context & Reliability — 10 questions)

Rest / review

Study context degradation in extended sessions, scratchpad files

Study /compact, subagent delegation, crash recovery manifests

Study human review: stratified sampling, field-level confidence

Study information provenance: claim-source mappings, temporal data

Study synthesis output: well-established vs contested, source characterizations

Practice Test 10 (Advanced Context — 10 questions)

Rest / review

Complete Exercise 1: Multi-Tool Agent with Escalation Logic

Complete Exercise 2: Claude Code Team Workflow Configuration

Complete Exercise 3: Structured Data Extraction Pipeline

Complete Exercise 4: Multi-Agent Research Pipeline

Full Practice Exam 1 (50 questions, all 6 scenarios)

Review all wrong answers, identify weak domains

Rest / review weak areas

Targeted review of weakest domain

Targeted review of second weakest domain

Full Practice Exam 2 (50 questions, all 6 scenarios)

Review wrong answers, fill gaps

Full Practice Exam 3 (50 questions, timed)

Light review of key concepts, anti-patterns, and gotchas

Exam day!