Ace the Panaversity Digital FTE Quiz 2026! Get study material, question answers, and proven strategies to pass easily. Start preparing today! Panaversity Digital FTE Quiz preparation made easy. Explore detailed question breakdowns, answers, and expert tips for better results.
Panaversity Digital FTE Quiz Preparation: Questions, Answers & Strategies
Chapter 1: The AI Agent Factory Paradigm Quiz Panaversity
Test your understanding of the foundational concepts that define AI-Driven Development and the Digital FTE vision. This assessment covers all 9 lessons in Chapter 1.
Perfect 👍 Here’s the full Question + Correct Answer list for all 30 questions, clean and ready for studying — no explanations, just Q&A:
1. In the Agent Maturity Model, what is the primary purpose of the Incubator stage (General Agents)?
✅ Exploration, discovery, and prototyping
2. How does MCP (Model Context Protocol) function as the ‘USB cable for AI’?
✅ It provides a standardized protocol so any MCP-compatible agent can use any MCP server
3. What is a Digital FTE?
✅ An autonomous AI agent executing the COMPLETE output of a human employee, focused on outcomes not tasks
4. In the ‘Productivity Trap vs Ownership Model’ story, why did Sarah get displaced while Marcus succeeded?
✅ Sarah used AI for productivity; Marcus built a Digital FTE encoding his expertise as a product he owns
5. Which tasks do orchestrators focus on rather than typists?
✅ Specification writing, requirement gathering, and validation of AI-generated work
6. When should you use General Agents according to the Agent Factory paradigm?
✅ When you’re not sure what the solution should look like
7. What is the practical implication of probabilistic LLM outputs for software development?
✅ Validation and testing become essential since outputs vary
8. What are the six phases of the SDD workflow?
✅ Specify, Clarify, Plan, Tasks, Implement, Validate
9. When is ‘Vibe Coding’ appropriate vs when is SDD essential?
✅ Vibe coding works for learning/throwaway code; SDD is essential for production features, security-critical, or AI-assisted development
10. What is an ‘M-Shaped Developer’ and why was it nearly impossible before AI?
✅ A developer with deep expertise in 2–4 complementary domains, enabled by AI handling cognitive load across areas
11. What is the OODA loop?
✅ A reasoning framework with Observe, Orient, Decide, Act—used by AI agents to process information and take action
12. What defines AI-Driven Development (AIDD)?
✅ A specification-first methodology transforming developers into specification engineers and architects
13. Why is partial adoption of AIDD pillars (e.g., 6 of 9) less effective than complete adoption?
✅ Partial adoption creates gaps; pillars multiply effects exponentially when combined, not just add linearly
14. What is the ‘Shadow Mode’ deployment strategy for high-risk domains?
✅ Agent runs and generates recommendations while humans make all final decisions, logging everything for comparison
15. In the ‘Snakes and Ladders’ framework, which layer should third-party developers avoid competing in?
✅ Layer 1: Consumer AI Backbone (OpenAI vs Google war)
16. What is the economic advantage of Digital FTE labor over human labor for customer support?
✅ Digital FTEs cost ~$3/ticket vs ~$150/ticket for humans, with 24/7 availability
17. Which of the following is concrete evidence that AI coding capability reached production quality in 2024–2025?
✅ OpenAI achieved a perfect score solving all 12 problems at the ICPC World Finals
18. Why is LLM output probabilistic rather than deterministic?
✅ Models sample from probability distributions, producing different outputs from identical inputs
19. According to Chapter 1, what is the fundamental choice developers face?
✅ Path A (treat AI as faster keyboard, vibe code) vs Path B (master Agent Factory paradigm, build Digital FTEs)
20. Why does ‘AI amplifies your habits’ matter for development methodology?
✅ Spec-driven development becomes MORE critical because AI amplifies both good discipline and bad habits
21. According to the 2025 Stack Overflow Developer Survey, what percentage of professional developers use or plan to use AI coding tools?
✅ 84%
22. What is the key difference between MCP and Agent Skills?
✅ MCP provides connectivity (how agents talk to tools); Skills provide expertise (what agents know how to do)
23. In MCP’s three primitives, what is the difference between Resources and Tools?
✅ Resources provide data to read (agent’s ‘eyes’); Tools execute actions (agent’s ‘hands’)
24. What is ‘The Moat’ in Digital FTE positioning?
✅ The 20% of nuanced, experience-based insights that generic AI cannot replicate
25. When would a ‘Success Fee’ model be better than ‘Subscription’ for Digital FTE pricing?
✅ When outcomes are easily measurable and clients are skeptical (‘prove it first’)
26. What is the primary advantage of a modular, three-layer AI stack compared to monolithic tool ecosystems?
✅ It prevents vendor lock-in and enables faster evolution by composing independent layers
27. What is ‘Progressive Disclosure’ in the Agent Skills standard?
✅ Loading only skill metadata at startup (~100 tokens), full instructions when activated (<5k), resources on-demand
28. What happens when an LLM’s context window fills up during a long conversation?
✅ Early messages get truncated, losing information
29. What does ‘AI is an amplifier’ mean for the choice between Vibe Coding and SDD?
✅ AI amplifies good habits (SDD) AND bad habits (Vibe Coding)—discipline matters MORE with AI, not less
30. In the Modern AI Stack, what role do AI-First IDEs (Layer 2) play?
✅ They act as context orchestrators, intelligently selecting relevant code for models
1. What are the four elements of a complete specification in SDD?
Answer: Intent, Success Criteria, Constraints, Non-Goals
Explanation: Defines purpose, success measures, limits, and what is explicitly excluded.
2. What is the OODA loop?
Answer: Observe, Orient, Decide, Act—reasoning framework for AI agents to process and act.
Explanation: Military strategy framework guiding continuous AI decision-making.
3. What creates the ‘virtuous cycle’ in the Agent Factory paradigm?
Answer: Clear specs → precise execution → reliable agents → Digital FTEs → multiplied capacity → larger problems → better specs.
Explanation: Each step builds on the previous, compounding capabilities.
4. Why is LLM output probabilistic rather than deterministic?
Answer: Models sample from probability distributions, so outputs vary even for identical inputs.
Explanation: Results differ slightly due to probabilistic nature.
5. What are the six phases of the SDD workflow?
Answer: Specify, Clarify, Plan, Tasks, Implement, Validate
Explanation: Defines steps from defining specs to validating final output.
6. What is the key difference between MCP and Agent Skills?
Answer: MCP connects agents to tools; Skills encode what agents know how to do.
Explanation: MCP = connectivity; Skills = expertise.
7. What is the core equation of Spec-Driven Development?
Answer: Vague Idea + AI = 5+ iterations; Clear Specification + AI = 1-2 iterations
Explanation: Clear specs reduce iteration cycles drastically.
8. What distinguishes Generation 4 AI tools from Generation 3?
Answer: Gen 4 agents execute autonomously with multi-turn capability; Gen 3 needed step-by-step approval.
Explanation: Gen 4 automates full workflows without human steps.
9. What is the primary advantage of a modular, three-layer AI stack compared to monolithic ecosystems?
Answer: Prevents vendor lock-in and enables faster evolution by composing independent layers.
Explanation: Modular design allows swapping and evolving components independently.
10. What distinguishes goose from Claude Code?
Answer: goose is open-source (Apache 2.0) with visible source code; Claude Code is proprietary.
Explanation: goose can be studied and adapted; Claude Code is closed-source.
11. Which approach helps manage the ‘context is limited’ constraint in LLMs?
Answer: Reference file paths rather than pasting entire contents.
Explanation: Saves token budget by referencing instead of duplicating info.
12. Why does ‘Markdown as Programming Language’ enable new development patterns?
Answer: Markdown specs become executable ‘source code’ that AI agents read and implement.
Explanation: Allows human-readable specs directly executable by AI.
13. What are the Five Powers that enable autonomous agents?
Answer: See, Hear, Reason, Act, Remember
Explanation: Combined sensory and cognitive capabilities enable autonomy.
14. What is ‘Progressive Disclosure’ in the Agent Skills standard?
Answer: Load only metadata initially; full instructions and resources load on demand.
Explanation: Reduces token usage and speeds agent startup.
15. Which is concrete evidence AI coding reached production quality in 2024-2025?
Answer: OpenAI achieved a perfect score solving all 12 problems at the ICPC World Finals.
Explanation: Demonstrates AI solving complex algorithmic problems.
16. In the ‘Snakes and Ladders’ framework, which layer should third-party developers avoid competing in?
Answer: Layer 1: Consumer AI Backbone (OpenAI vs Google war)
Explanation: Only dominant players survive here; avoid this brutal competition.
17. What is the economic advantage of Digital FTE labor over human labor for customer support?
Answer: Digital FTEs cost ~$3/ticket vs ~$150/ticket for humans, with 24/7 availability.
Explanation: Much cheaper and always available.
18. What is ‘The Moat’ in Digital FTE positioning?
Answer: The 20% of nuanced, experience-based insights generic AI cannot replicate.
Explanation: Human judgment and expertise remain critical.
19. According to Chapter 1, what is the fundamental choice developers face?
Answer: Path A (AI as faster keyboard) vs Path B (master Agent Factory paradigm, build Digital FTEs).
Explanation: Choose between reactive coding or systematic AI-driven scaling.
20. In the Agent Maturity Model, what is the primary purpose of the Incubator stage?
Answer: Exploration, discovery, and prototyping.
Explanation: General agents used for flexible reasoning, not production.
21. In the Modern AI Stack, what role do AI-First IDEs play?
Answer: Context orchestrators, intelligently selecting relevant code for models.
Explanation: Manage context to improve AI coding.
22. Why is partial adoption of AIDD pillars less effective than complete adoption?
Answer: Pillars multiply effects exponentially when combined; partial creates gaps.
Explanation: Full integration needed for maximum benefit.
23. Why does ‘AI amplifies your habits’ matter for development methodology?
Answer: Spec-driven development is more critical because AI amplifies both good and bad habits.
Explanation: Discipline must increase to avoid amplified chaos.
24. In the AI-transformed SDLC, what does the developer do during Coding?
Answer: Validate AI-generated code against specs and security requirements.
Explanation: Shift from coding to quality validation.
25. When is a ‘Success Fee’ model better than ‘Subscription’ for Digital FTE pricing?
Answer: When outcomes are measurable and clients are skeptical (‘prove it first’).
Explanation: Aligns cost with verified results.
26. What is the practical implication of probabilistic LLM outputs for software development?
Answer: Validation and testing become essential since outputs vary.
Explanation: Ensures reliability despite output variability.
27. What happens when an LLM’s context window fills up during a long conversation?
Answer: Early messages get truncated, losing information.
Explanation: Important to manage context window strategically.
28. When should you use General Agents according to Agent Factory paradigm?
Answer: When requirements are unclear, changing, or novel.
Explanation: General agents good for exploration and prototyping.
29. What is the primary business benefit of AAIF for building Digital FTEs?
Answer: Ensures Digital FTEs are portable and work across platforms, not locked.
Explanation: Avoids vendor lock-in with open standards.
30. When a developer says ‘It remembers my coding style,’ what’s actually happening?
Answer: No true memory; style is maintained by reinjecting context via style guides (AGENTS.md).
Explanation: Persistent style preferences encoded externally.
If you want, I can make this into a neat PDF or flashcards, or explain any point in detail!
Chapter 2: Markdown – Writing Instructions Quiz Panaversity
Here’s your clean, complete Q + A list for all 18 questions:
1. You’re specifying a feature where the implementation approach depends on business constraints explained in prose paragraphs. AI implements a technically correct solution that violates business constraints. What structural addition might have prevented this?
✅ Add a ‘## Constraints’ heading with a bulleted list identifying business constraints
2. A developer uses six levels of heading hierarchy in a specification. Another developer uses only two levels. What does heading depth primarily communicate about the specification?
✅ Heading depth reflects the complexity and granularity of the specification’s organization
3. You include a code block showing error handling but don’t specify which errors should be caught. AI implements generic error handling. What additional context would have improved clarity?
✅ Add comments within the code block or surrounding text specifying which errors
4. A specification includes a code block without a language tag. What limitation does this create?
✅ AI loses context about language for interpreting syntax and providing implementation guidance
5. Your specification has a list of user stories. You use ordered lists because you have five stories. During implementation, AI treats these as sequential phases. What caused this misunderstanding?
✅ Ordered lists communicate sequence; if independent, unordered lists avoid implying sequence
6. You specify API endpoints in an unordered list, then realize some endpoints depend on others being implemented first. What change would communicate this dependency to AI?
✅ Convert to an ordered list showing the implementation sequence for endpoints
7. You’re specifying a feature with multiple independent capabilities that can be implemented in any order. Which list type best communicates this to AI?
✅ Unordered list to indicate these are independent items without required sequence
8. In a specification, you place ‘## Success Criteria’ before describing the feature itself. Why might this heading placement improve AI’s implementation?
✅ AI reads specifications top-to-bottom, so early success criteria can guide all subsequent implementation decisions
9. In a specification, you describe a complex user workflow with multiple decision points. Which markdown approach will make it easiest for AI to understand the sequence and dependencies?
✅ Use ordered lists for sequential steps and nested lists for decision branches
10. You ask AI to build a login feature by describing it conversationally: “Make a login page that looks nice and works well.” What fundamental problem does this approach create?
✅ The specification lacks structural clarity about specific requirements and success criteria
11. Two specifications describe the same feature: one uses descriptive headings (“User Authentication System”), the other uses vague headings (“Feature 1”). How will this difference affect AI’s implementation?
✅ Descriptive headings help AI maintain context and make decisions consistent with the feature’s purpose
12. Your specification shows expected API response using a code block. AI implements the feature but returns a different JSON structure. What aspect of your code block likely caused this misunderstanding?
✅ The code block lacked context about whether it was complete or partial
13. Your specification includes a numbered list of features, each containing a bulleted list of requirements. What hierarchical relationship does this communicate to AI?
✅ Each feature contains multiple requirements; features are sequential, requirements are order-independent
14. You notice that AI misunderstood your specification and implemented the wrong feature. What aspect of markdown structure would have most likely prevented this misunderstanding?
✅ Using clearer headings to separate distinct features and establish scope boundaries
15. A team debates whether to write specifications in Word documents or markdown files. What is the strongest technical argument for choosing markdown for AI collaboration?
✅ Markdown is plain text AI can parse without proprietary format conversion
16. A specification shows database schema using a fenced code block tagged as “`sql. AI implements the schema but uses slightly different syntax. What does the language tag help AI understand?
✅ The language tag indicates SQL is the target language and informs syntax expectations and patterns
17. When using “specification by example” in markdown, what are you demonstrating to AI, and why is this more effective than pure description?
✅ You show concrete expected inputs and outputs rather than just describing behavior
18. A specification has dozens of level 2 headings with no level 1 heading at the top. What organizational problem does this create for AI parsing?
✅ The document lacks a clear top-level scope that establishes context for all subsections
Enroll in the free ADBI Cybersecurity Course 2026. Earn a verified digital certificate in 90 minutes. Learn data protection, phishing prevention, and online safety. In an era where digital threats are evolving daily, foundational knowledge of data protection is no longer optional—it’s a necessity. The ADBI Cybersecurity Course 2026, offered by the Asian Development Bank…
Continue Reading Free ADBI Cybersecurity Certificate 2026: 90-Min Online Course
Digital marketing has become an essential skill for anyone looking to promote products or services online. In this first day of our practical digital marketing series, we will focus on understanding the basics of digital marketing and performing keyword research. By the end of this lesson, you will have a strong foundation to create your…
Digital platforms that combine simplicity, effectiveness, and rich experiences always stand out. One such platform grabbing the attention of many Indian users is Dreamexch, or as some call it, “dream exchange.” Dreamexchange is more than just another website or app – it is an all-in-one platform for creating and connecting live, digital content through entertainment…
Continue Reading Dreamexchange 2026 Review: UX, Features & What Sets It Apart
For most digital marketing agencies, growth hits a ceiling when clients begin to demand full-scale SEO services. In fact, building an in-house SEO team will require great investment in talent, tools, and ongoing training, which can be expensive and time-consuming for a small or mid-sized agency. That’s where the reseller SEO companies come in. The…
Continue Reading Top 8 Reseller SEO Companies You Can Count On for Growth
Enroll in this free Claude Code course to learn how to use AI coding assistants effectively. Understand Plan Mode, Thinking Mode, hooks, permissions, and custom commands. Get a certificate upon completion! Perfect for developers wanting to enhance their AI-powered coding skills. In this free course, you’ll explore: ✅ Plan Mode vs Thinking Mode✅ How to…
Continue Reading Free Claude Code Course with Certificate 2026 | Master AI Coding Assistant
Learn how to create a complete command-line todo application using Python 3.13+, Claude Code AI assistance, and Spec-Kit Plus specification-driven development. Perfect for beginners Basic Level FunctionalityObjective: Build a command-line todo application that stores tasks in memory using ClaudeCode and Spec-Kit Plus.Requirements Implement all 5 Basic Level features (Add, Delete, Update, View, Mark Complete) Use spec-driven…
Continue Reading The Evolution of Todo – Mastering Spec-Driven Development & Cloud Native AI
Question 1
Your answer: Code blocks must always appear in output-input-process order for AI parsing
Correct answer: The input-process-output sequence shows complete behavior flow, demonstrating transformation from input to output
Explanation:
The sequence input → process → output gives a clear picture of how data is transformed step-by-step. This helps AI understand the feature’s intended behavior at each stage. The code blocks serve as examples, not exact code to copy. Multiple smaller blocks with clear purposes are better than one big block. While order is flexible, input-process-output is the most intuitive and informative sequence.
Question 2
Your answer: The language tag forces AI to use exactly the syntax shown without any modifications
Correct answer: The language tag indicates SQL is the target language and informs syntax expectations and patterns
Explanation:
The “`sql tag tells AI that the code is SQL, guiding it to use appropriate syntax and conventions. However, AI may adapt syntax based on the specific database or context. The tag is a guideline, not a strict enforcement of exact syntax replication.
Question 3
Your answer: Use bold text for the first word of each phase description
Correct answer: Add an ordered list showing the three phases explicitly numbered
Explanation:
An ordered list with numbers clearly communicates sequence and dependencies, which AI can parse unambiguously. Bold text provides emphasis but does not convey structural order or hierarchy to AI.
Question 4
Your answer: AI will interpret ‘looks nice’ as a command to use CSS
Correct answer: The specification lacks structural clarity about specific requirements and success criteria
Explanation:
Vague, conversational descriptions lack the structure needed to define clear requirements or measurable success criteria. Without concrete examples or criteria, AI cannot determine what exactly “looks nice” or “works well” means in practice.
Question 5
Your answer: Showing results isn’t sufficient; specify tests should be created with example structure
Correct answer: Showing expected results alone doesn’t explicitly require test creation; tests should be specified with example test structure
Explanation:
Expected results hint at what tests should check, but unless the spec explicitly requires test creation (with examples of test code), AI may not create tests. A complete test specification includes instructions and sample test code.
Question 6
Your answer: No list at all, just paragraphs describing each capability
Correct answer: Unordered list to indicate these are independent items without required sequence
Explanation:
Unordered lists (bullets) clearly indicate independent capabilities without implying order, which helps AI understand they can be implemented in any sequence.
Question 7
Your answer: Images in markdown are automatically converted to code by AI without human interpretation needed
Correct answer: AI can analyze images when explicitly directed but may not automatically incorporate image content into implementation
Explanation:
AI does not automatically translate images into implementation. To ensure AI uses image details, descriptive text should accompany images. Images alone serve as visual references.
Question 8
Your answer: Ordered lists cannot be used for items without dependencies in markdown
Correct answer: AI implements requirements sequentially even when parallel implementation would be faster
Explanation:
Ordered lists imply sequence and dependencies. Using them for independent requirements can cause AI to unnecessarily implement features sequentially rather than in parallel.
Question 9
Your answer: Italic and bold have identical semantic meaning for AI parsing based on preference
Correct answer: Bold provides stronger emphasis; AI may interpret it as higher priority
Explanation:
Bold text signals stronger emphasis and can imply higher priority to AI, whereas italic is weaker. Proper priority is best communicated with structural elements like headings.
Question 10
Your answer: Markdown’s modular syntax allows incremental addition of structure without reformatting everything
Correct answer: Markdown’s modular syntax supports adding structure incrementally without reformatting existing content
Explanation:
Markdown lets you build up specifications step-by-step—adding headings, lists, and code blocks—without breaking previous content, enabling smooth iterative development.
Question 11
Your answer: The code block showing data structure, providing concrete technical specifications and patterns
Correct answer: Code blocks showing data structures most directly influence code structure
Explanation:
Code blocks give AI explicit templates for data structures or APIs, which have a direct impact on the implemented code, unlike mockups or business rules that guide UI or logic.
Question 12
Your answer: Intent Layer, because markdown structures human requirements for AI interpretation
Correct answer: Markdown functions in the Intent Layer, structuring human intent for AI interpretation
Explanation:
Markdown captures what needs to be done without specifying how, bridging human ideas and AI implementation in the Intent Layer.
Question 13
Your answer: AI always automatically fetches and reads all linked documentation before implementation
Correct answer: AI cannot access external links during implementation, so linked resources don’t directly inform implementation
Explanation:
AI works only with the provided text. External links serve as references for humans, not direct inputs for AI, unless their content is included explicitly.
Question 14
Your answer: Use fenced code blocks with language tags showing example input/output or behavior
Correct answer: Use fenced code blocks with language tags to specify expected behavior examples
Explanation:
Code blocks with proper language tags provide clear, formatted examples of expected input/output, making specifications unambiguous and easier for AI to implement.
Question 15
Your answer: The paragraph-based specification will be faster for AI to process completely
Correct answer: Structured specifications enable AI to generate more organized and accurate code
Explanation:
Headings, lists, and code blocks help AI parse requirements and relationships better than plain paragraphs, leading to better implementation quality.
Question 16
Your answer: Use level 1 for the feature, level 2 for components, level 3 for sub-features
Correct answer: Use hierarchical headings (#, ##, ###) to reflect feature, components, and sub-features
Explanation:
Proper heading levels clarify the structure and relationships among parts of the feature, which AI uses to understand scope and hierarchy.
Question 17
Your answer: Italic formatting should automatically signal placeholders to all AI systems
Correct answer: Use inline code with angle brackets like <database_name> and explanatory text for placeholders
Explanation:
Italics are not a reliable placeholder indicator. Using code formatting and explicit markers like < > plus explanations makes it clear which parts are placeholders.
Question 18
Your answer: Unordered lists should never be used in specifications for AI implementation
Correct answer: The requirements had sequential dependencies that should have been expressed using ordered lists
Explanation:
If dependencies exist, unordered lists mislead AI into treating items as independent. Ordered lists properly communicate sequence and dependencies.
If you want, I can help you write or improve specifications based on these principles!
Chapter 3: Claude Code and Cowork Quiz Panaversity
Test your understanding of Claude Code and Claude Cowork’s architecture, extensibility features, and workflow patterns. This assessment covers installation, configuration, MCP integration, subagents, skills, hooks, settings hierarchy, plugin architecture, browser integration, connectors, and built-in document Skills.
Perfect 👍 Here’s your Chapter 3: Claude Code and Cowork Quiz — complete with all 15 questions and their correct answers only (no explanations), formatted for easy studying:
1. You’re adding testing requirements to CLAUDE.md. Which directive would most effectively ensure Claude Code writes testable code?
✅ All functions must be pure when possible; provide test example patterns
2. You need to write Python scripts to analyze data, then create a formatted Word report with the results. Which approach is most appropriate?
✅ Use Claude Code for the Python scripts, then switch to Claude Cowork for the Word report
3. You’ve created a subagent for code review tasks. During testing, it generates excellent feedback but takes much longer than expected. Which subagent configuration would most directly address this performance issue?
✅ Configure subagent to use Haiku model instead of Sonnet or Opus
4. You’re developing a Claude Code plugin that adds Git workflow automation. The plugin needs to read repository history and create commits. Which integration point would provide the most appropriate access to these capabilities?
✅ Plugin should expose Git operations as MCP tools Claude Code invokes
5. Why can’t a developer just copy code from ChatGPT into their editor instead of using Claude Code?
✅ Copy-paste workflows break iterative refinement loops where AI validates its own changes
6. Your team lacks programming experience but can write documentation and YAML config. Which Claude Code extensibility mechanisms remain accessible?
✅ CLAUDE.md, settings, agent skills using markdown prompt patterns accessible
7. After installing Claude Code globally, you get “command not found.” What does this reveal about npm global installation and PATH?
✅ npm global installs require the npm bin directory in PATH variable
8. What’s the actual difference between Claude Code and Claude Cowork?
✅ They’re the same AI with different interfaces: terminal for developers (Code) and desktop for knowledge workers (Cowork)
9. You’ve created a skill for database schema design, but it ignores naming conventions defined in CLAUDE.md. What does this reveal?
✅ Skills should reference project context explicitly when conventions matter for output
10. Claude Code implemented a feature using a pattern inconsistent with team practices. What does this suggest?
✅ Agentic AI requires explicit project context to align actions with team conventions
11. Your MCP server exposes a ‘deploy_to_production’ tool that was triggered accidentally. What’s the best safeguard?
✅ Add required confirmation parameters to the deploy tool forcing explicit approval
12. How does Claude Code address the concern that “AI might make destructive changes”?
✅ Hooks allow validation logic to review and approve changes before execution
13. Your CLAUDE.md says “Always use async/await instead of promises,” but Claude Code still uses .then(). What does this reveal?
✅ CLAUDE.md requires clear context and specific examples to reliably guide behavior
14. After authenticating Claude Code, how can you verify the connection works before starting a complex task?
✅ Run ‘claude-code chat’ and send a simple test message
15. Your team uses CLAUDE.md, MCP tools, hooks, and agent skills. A new developer asks which to use for a specific task. What principle guides this decision?
✅ Each mechanism addresses different architectural concerns, so understand the problem first
Question 1
Correct
Principle: Each mechanism serves a specific role; choose based on problem domain.
Explanation: CLAUDE.md for context, MCP for tools, hooks for validation, skills for reasoning. Pick the right tool for the task.
Question 2
Incorrect
Correct Answer: CLAUDE.md context is automatically injected into every Claude Code conversation.
Explanation: CLAUDE.md ensures project context is always available to Claude Code, unlike manual wiki references.
Question 3
Incorrect
Correct Answer: Provide sensible defaults with progressive disclosure for advanced options.
Explanation: Make plugin easy to start using with defaults, reveal complexity only as needed.
Question 4
Incorrect
Correct Answer: Use Claude Code for Python scripts, then Claude Cowork for Word report.
Explanation: Use each tool for what it is optimized: coding vs document creation.
Question 5
Incorrect
Correct Answer: Pre-tool-call hook checking for database tool use and blocking during review tasks.
Explanation: Enforce constraints as early as possible (before tool call) to prevent unsafe operations.
Question 6
Incorrect
Correct Answer: Project-level settings override global settings for specificity.
Explanation: More specific context takes precedence over general defaults.
Question 7
Incorrect
Correct Answer: Subagents allow delegation to specialized agents with different tools.
Explanation: Use subagents to orchestrate complex workflows by dividing tasks.
Question 8
Correct
Best Practice: Make server port configurable with fallback to a random free port.
Explanation: Avoid port conflicts automatically with sensible defaults.
Question 9
Incorrect
Correct Answer: Documentation explaining component purposes, examples, and boundaries improves learnability.
Explanation: Teach mental models and concepts, not just APIs.
Question 10
Correct
Solution: Configure subagent to use faster Haiku model instead of Sonnet or Opus.
Explanation: Optimize model choice first before adding orchestration complexity.
Question 11
Correct
Integration Point: Expose Git operations as MCP tools Claude Code can invoke.
Explanation: Provide capabilities as tools; let Claude Code decide when/how to use them.
Question 12
Incorrect
Correct Answer: CLAUDE.md needs clear, specific instructions with examples to guide behavior.
Explanation: Vague rules don’t reliably change output; provide rationale and examples.
Question 13
Incorrect
Correct Answer: Local settings override project settings for specific developers.
Explanation: Settings cascade with most specific (local) having highest priority.
Question 14
Incorrect
Correct Answer: Create MCP server exposing database tools that use the internal package.
Explanation: Add project-specific capabilities as MCP tools for direct invocation.
Question 15
Correct
Insight: Orchestration requires explicit dependency management and error handling.
Explanation: Subagent workflows need intentional control flow to handle failures properly.
Chapter 4: The Seven Principles of General Agent Problem Solving Quiz Panaversity
Test your understanding of the seven principles that make agentic AI workflows effective: bash as universal interface, code as universal interface, verification as core step, small reversible decomposition, persisting state in files, constraints and safety, and observability.
Here’s your Chapter 4: The Seven Principles of General Agent Problem Solving Quiz — complete with all 15 questions and their correct answers only, formatted cleanly for review or study:
1. When debugging an AI session, what information is most valuable for understanding what went wrong?
✅ The activity log showing the sequence of actions taken
2. Why is it recommended to create a feature branch for AI-assisted work rather than working directly on main?
✅ Feature branches enable easy rollback if AI makes mistakes
3. Six months after choosing PostgreSQL over MongoDB, the team forgets why they made that decision. What document would have prevented this memory loss?
✅ Architecture Decision Record (ADR) explaining the choice
4. A team accepts AI-generated code without review because “the tests pass.” After deployment, they discover a security vulnerability. What verification principle did they violate?
✅ They skipped manual review which catches issues tests may miss
5. An AI tool offers to “auto-approve all commands” for faster workflow. When is this appropriate?
✅ When working in sandbox environment on trusted codebase
6. You ask an AI to “Add input validation.” It generates basic length checks, but you wanted business rule validation. What communication breakdown occurred?
✅ Natural language ambiguity prevented precise intent transmission
7. You ask AI to “refactor the auth system” and it immediately starts making changes. What’s missing from this workflow?
✅ The workflow lacks decomposition into small, verifiable steps
8. You’re explaining to a colleague why terminal access matters for AI. They say: “I can just copy-paste commands.” What limitation did they miss?
✅ Copy-paste breaks the OODA loop that enables autonomous iteration
9. You’re working on a critical system (payment processing). Which permission mode should you use?
✅ Confirming mode—approve all writes to maintain oversight
10. A team uses a prose specification, and AI implementations miss edge cases. What would improve alignment between specs and implementations?
✅ Include executable test cases as part of the specification
11. You’re debugging a production issue. Which three principles are most critical?
✅ Bash access, verification, and observability
12. You grant AI terminal access on a production server and it deletes important files. What safety principle was violated?
✅ Constraints and safety—AI shouldn’t have production access
13. An AI-generated function passes tests but fails under concurrent load due to a race condition. Why didn’t verification catch this?
✅ The race condition only occurs under concurrent load not captured by basic tests
14. You’re implementing a new feature. Which principles are most critical for this task type?
✅ All principles are relevant for feature implementation work
15. A developer re-explains project conventions in every new session. What would solve this repetition?
✅ Persist project context in CLAUDE.md that AI reads automatically
Question 1
Correct Answer: Bash access, verification, and observability
Key Principle: Debugging requires direct system access (bash), testing fixes (verification), and visibility into system behavior (observability).
Question 2
Correct Answer: Break the change into smaller pieces and verify each independently
Key Principle: Large changes must be decomposed into manageable, verifiable steps for effective review.
Question 3
Correct Answer: Architecture Decision Record (ADR) explaining the choice
Key Principle: Document design decisions with ADRs to preserve rationale for future reference.
Question 4
Correct Answer: All principles are relevant for feature implementation work
Key Principle: Feature work spans many principles—decomposition, verification, safety, observability, state persistence, bash access, and code as interface.
Question 5
Correct Answer: Feature branches enable easy rollback if AI makes mistakes
Key Principle: Use feature branches as reversible workspaces for AI-assisted development to isolate errors.
Question 6
Correct Answer: Identify which principle would provide the most value for their specific pain points
Key Principle: Start improvement by targeting the biggest workflow bottleneck rather than applying all principles at once.
Question 7
Correct Answer: The workflow lacks decomposition into small, verifiable steps
Key Principle: Break down large refactoring tasks into small steps for reviewability and rollback.
Question 8
Correct Answer: Constraints and safety—AI shouldn’t have production access
Key Principle: Isolate AI from production environments to prevent catastrophic errors.
Question 9
Correct Answer: Small, reversible decomposition with atomic commits
Key Principle: Make atomic commits to isolate and debug changes efficiently.
Question 10
Correct Answer: Copy-paste breaks the OODA loop that enables autonomous iteration
Key Principle: Terminal access lets AI autonomously observe, decide, and act, closing the feedback loop.
Question 11
Correct Answer: Provide specific code examples showing desired behavior
Key Principle: Use precise code examples to communicate intent and reduce ambiguity.
Question 12
Correct Answer: Natural language ambiguity prevented precise intent transmission
Key Principle: Ambiguous instructions need clarification with examples or detailed specs.
Question 13
Correct Answer: Terminal access allows the AI to read test files, run tests, and observe actual output
Key Principle: Terminal access enables AI to actively debug by running and analyzing tests.
Question 14
Correct Answer: Observability—AI should show what it’s doing as it works
Key Principle: Visibility into AI actions prevents unnoticed unwanted changes.
Question 15
Correct Answer: They skipped manual review which catches issues tests may miss
Key Principle: Tests are necessary but not sufficient; manual review catches hidden risks like security.