In-Scope vs Out-of-Scope

In-Scope Topics

• Agentic loop implementation: stop_reason control flow, tool result handling, loop termination
• Multi-agent orchestration: coordinator-subagent patterns, task decomposition, parallel execution, iterative refinement loops
• Subagent context management: explicit context passing, structured state persistence, crash recovery via manifests
• Tool interface design: effective descriptions, split vs. consolidate, naming to reduce ambiguity
• MCP tool/resource design: resources for catalogs, tools for actions, description quality
• MCP server configuration: project vs. user scope, env var expansion, multi-server simultaneous access
• Error handling/propagation: structured responses, transient/business/permission errors, local recovery before escalation
• Escalation decision-making: explicit criteria, honoring customer preference, policy-gap identification
• CLAUDE.md configuration: hierarchy, @import, .claude/rules/ glob patterns
• Custom commands/skills: project vs. user scope, context: fork, allowed-tools, argument-hint
• Plan mode vs. direct execution: complexity assessment
• Iterative refinement: I/O examples, TDD iteration, interview pattern, sequential vs. parallel fixes
• Structured output via tool_use: schema design, tool_choice, nullable fields
• Few-shot prompting: ambiguous scenarios, format consistency, FP reduction
• Batch processing: Batches API fit, latency tolerance, custom_id failure handling
• Context window optimization: trimming verbose outputs, structured fact extraction, position-aware ordering
• Human review workflows: confidence calibration, stratified sampling, accuracy segmentation
• Information provenance: claim-source mappings, temporal data, conflict annotation, coverage gaps

Explicitly Out-of-Scope

• Fine-tuning / training custom models
• API authentication, billing, account management
• Deep language/framework implementation details (beyond tool/schema config needs)
• Deploying/hosting MCP servers (infra, networking, container orchestration)
• Claude's internal architecture, training process, model weights
• Constitutional AI, RLHF, safety training methodology
• Embedding models / vector DB implementation
• Computer use (browser/desktop automation)
• Vision/image analysis
• Streaming API / server-sent events
• Rate limiting, quotas, pricing calculations
• OAuth, API key rotation, auth protocol details
• Cloud provider specifics (AWS/GCP/Azure)
• Performance benchmarking / model comparison metrics
• Prompt caching implementation details (just know it exists)
• Token counting algorithms / tokenization specifics

Takeaway: if a question drifts into infra/ops/ML-training territory, that's a signal you may be over-thinking it — the exam stays at the architecture/config/prompt-design layer.