diff --git a/CHANGELOG.md b/CHANGELOG.md
index 6bda841..09a4657 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -6,6 +6,30 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/).
 
 ## [Unreleased]
 
+## [3.20.4] - 2026-01-31
+
+### Added
+
+- **30 New Quiz Questions** (227 → 257 total) across 11 categories
+  - **Advanced Patterns** (+8): Mechanic Stacking, Permutation Frameworks, "You Are the Main Thread", Task Lists as Diagnostic, Anti-hallucination occurrence rule, Multi-Agent PR Review, Comprehension Debt, CLAUDE.md compounding memory
+  - **MCP Servers** (+3): MCP Apps (SEP-1865), `auto:N` lazy loading, Semgrep top score (9.0/10)
+  - **Architecture** (+3): Tasks API (replaces TodoWrite), Tasks API N+1 overhead, TeammateTool experimental status
+  - **Reference** (+3): `--from-pr` flag, `$ARGUMENTS` bracket syntax breaking change, Myths vs Reality
+  - **Hooks** (+2): Async hook `async: true` config, async hook limitations
+  - **Learning with AI** (+2): Addy Osmani's 80% Problem failure modes, vibe coding context overload symptoms
+  - **Security** (+2): Docker sandbox isolation, GitHub Issue Auto-Creation Bug (#13797)
+  - **AI Ecosystem** (+3): LM Studio bridge cost savings, external orchestrators (Gas Town/multiclaude/agent-chat), skeleton project audit areas
+  - **Memory & Settings** (+2): 8 verification domains, Fresh Context Pattern
+  - **Agents** (+1): AGENTS.md vs Skills invocation reliability (100% vs 53-79%)
+  - **Privacy** (+1): Co-Authored-By vs Assisted-By traceability
+  - Difficulty distribution: 4 junior, 4 intermediate, 14 senior, 8 power
+  - Also fixed pre-existing "14 categories" → "15 categories" in landing quiz page
+
+### Changed
+
+- **README.md**: Quiz badge updated (227 → 257), quiz section updated
+- **Landing site**: All quiz counts updated (index.html, quiz/index.html, learning/index.html, CLAUDE.md)
+
 ## [3.20.3] - 2026-01-31
 
 ### Added
diff --git a/CLAUDE.md b/CLAUDE.md
index fc59b57..a6d1cb6 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -279,7 +279,7 @@ Ce guide fait partie d'un écosystème de 4 repositories interconnectés, sépar
 | Aspect | Détails |
 |--------|---------|
 | **Local** | `/Users/florianbruniaux/Sites/perso/claude-code-ultimate-guide-landing/` |
-| **Contenu** | Page marketing, badges, FAQ, quiz (227 questions) |
+| **Contenu** | Page marketing, badges, FAQ, quiz (257 questions) |
 | **Sync avec** | Guide principal (version, templates, guide lines) |
 
 ### 4. Cowork Landing Site
diff --git a/README.md b/README.md
index 86556a3..05a1b9e 100644
--- a/README.md
+++ b/README.md
@@ -6,7 +6,7 @@
 
 <p align="center">
   <a href="https://github.com/FlorianBruniaux/claude-code-ultimate-guide/stargazers"><img src="https://img.shields.io/github/stars/FlorianBruniaux/claude-code-ultimate-guide?style=for-the-badge" alt="Stars"/></a>
-  <a href="./quiz/"><img src="https://img.shields.io/badge/Quiz-227_questions-orange?style=for-the-badge" alt="Quiz"/></a>
+  <a href="./quiz/"><img src="https://img.shields.io/badge/Quiz-257_questions-orange?style=for-the-badge" alt="Quiz"/></a>
   <a href="./examples/"><img src="https://img.shields.io/badge/Templates-86-green?style=for-the-badge" alt="Templates"/></a>
 </p>
 
@@ -68,7 +68,7 @@ graph LR
 
     root --> guide[📖 guide/<br/>16K lines]
     root --> examples[📋 examples/<br/>86 templates]
-    root --> quiz[🧠 quiz/<br/>227 questions]
+    root --> quiz[🧠 quiz/<br/>257 questions]
     root --> tools[🔧 tools/<br/>utils]
     root --> machine[🤖 machine-readable/<br/>AI index]
     root --> docs[📚 docs/<br/>25 evaluations]
@@ -102,7 +102,7 @@ graph LR
 │  ├─ skills/             1 meta-skill (Claudeception)
 │  └─ scripts/            Utility scripts (audit, search)
 │
-├─ 🧠 quiz/               227 Questions
+├─ 🧠 quiz/               257 Questions
 │  ├─ 9 categories        Setup, Agents, MCP, Trust, Advanced...
 │  ├─ 4 profiles          Junior, Senior, Power User, PM
 │  └─ Instant feedback    Doc links + score tracking
@@ -132,7 +132,7 @@ We explain **concepts first**, not just configs:
 - [Trade-offs](./guide/ultimate-guide.md#when-to-use-what) — When to use agents vs skills vs commands
 - [Pitfalls](./guide/ultimate-guide.md#common-mistakes) — Common mistakes and solutions
 
-### 📝 227-Question Quiz (Unique in Ecosystem)
+### 📝 257-Question Quiz (Unique in Ecosystem)
 
 **Only comprehensive assessment available** — test your understanding across 9 categories:
 - Setup & Configuration
@@ -407,7 +407,7 @@ Claude Code sends your prompts, file contents, and MCP results to Anthropic serv
 </details>
 
 <details>
-<summary><strong>Knowledge Quiz</strong> (227 questions)</summary>
+<summary><strong>Knowledge Quiz</strong> (257 questions)</summary>
 
 Test your Claude Code knowledge with an interactive CLI quiz covering all guide sections.
 
@@ -478,12 +478,12 @@ See [CONTRIBUTING.md](./CONTRIBUTING.md) for guidelines.
 
 ### Tools
 - [Ask Zread](https://zread.ai/FlorianBruniaux/claude-code-ultimate-guide) — Ask questions about this guide
-- [Interactive Quiz](./quiz/) — 227 questions
+- [Interactive Quiz](./quiz/) — 257 questions
 - [Landing Site](https://florianbruniaux.github.io/claude-code-ultimate-guide-landing/) — Visual navigation
 
 ---
 
-*Version 3.20.3 | January 2026 | Crafted with Claude*
+*Version 3.20.4 | January 2026 | Crafted with Claude*
 
 <!-- SEO Keywords -->
 <!-- claude code, claude code tutorial, anthropic cli, ai coding assistant, claude code mcp,
diff --git a/VERSION b/VERSION
index 7a5205a..c253e50 100644
--- a/VERSION
+++ b/VERSION
@@ -1 +1 @@
-3.20.3
+3.20.4
diff --git a/guide/ai-ecosystem.md b/guide/ai-ecosystem.md
index df051b7..e841f34 100644
--- a/guide/ai-ecosystem.md
+++ b/guide/ai-ecosystem.md
@@ -1800,7 +1800,7 @@ An **interactive comparison matrix** of 23 AI coding agents across 11 technical
 | "Which agents exist?" | "How to use Claude Code effectively?" |
 | Feature comparison (11 criteria) | Workflows, architecture, TDD/SDD methodologies |
 | 23 agents × shallow | 1 agent × deep (11K lines) |
-| Technical specs | Practical templates (66+), quiz (227 questions) |
+| Technical specs | Practical templates (66+), quiz (257 questions) |
 
 **Use case**: Use Matrix to **discover and compare** → Choose Claude Code → Use this guide to **master it**.
 
diff --git a/guide/cheatsheet.md b/guide/cheatsheet.md
index 8a78a9e..59f4daa 100644
--- a/guide/cheatsheet.md
+++ b/guide/cheatsheet.md
@@ -6,7 +6,7 @@
 
 **Written with**: Claude (Anthropic)
 
-**Version**: 3.20.3 | **Last Updated**: January 2026
+**Version**: 3.20.4 | **Last Updated**: January 2026
 
 ---
 
@@ -484,4 +484,4 @@ where.exe claude; claude doctor; claude mcp list
 
 **Author**: Florian BRUNIAUX | [@Méthode Aristote](https://methode-aristote.fr) | Written with Claude
 
-*Last updated: January 2026 | Version 3.20.3*
+*Last updated: January 2026 | Version 3.20.4*
diff --git a/guide/ultimate-guide.md b/guide/ultimate-guide.md
index 94f2aa2..998e621 100644
--- a/guide/ultimate-guide.md
+++ b/guide/ultimate-guide.md
@@ -10,7 +10,7 @@
 
 **Last updated**: January 2026
 
-**Version**: 3.20.3
+**Version**: 3.20.4
 
 ---
 
@@ -16290,4 +16290,4 @@ We'll evaluate and add it to this section if it meets quality criteria.
 
 **Contributions**: Issues and PRs welcome.
 
-**Last updated**: January 2026 | **Version**: 3.20.3
+**Last updated**: January 2026 | **Version**: 3.20.4
diff --git a/machine-readable/reference.yaml b/machine-readable/reference.yaml
index 95861bb..ff55d1f 100644
--- a/machine-readable/reference.yaml
+++ b/machine-readable/reference.yaml
@@ -3,7 +3,7 @@
 # Source: guide/ultimate-guide.md
 # Purpose: Condensed index for LLMs to quickly answer user questions about Claude Code
 
-version: "3.20.3"
+version: "3.20.4"
 updated: "2026-01-31"
 
 # ════════════════════════════════════════════════════════════════
@@ -782,7 +782,7 @@ rules:
 ecosystem:
   this_guide:
     focus: "Education - Learn & master"
-    unique: ["architecture docs", "TDD/SDD methodologies", "227-question quiz", "YAML index"]
+    unique: ["architecture docs", "TDD/SDD methodologies", "257-question quiz", "YAML index"]
   complementary:
     claude_code_templates:
       url: "github.com/davila7/claude-code-templates"
@@ -855,7 +855,7 @@ ecosystem:
       local: "/Users/florianbruniaux/Sites/perso/claude-code-ultimate-guide-landing/"
       purpose: "Marketing site for Claude Code guide"
       audience: "Developers discovering the guide"
-      features: ["badges (version, templates, lines)", "quiz (227 questions)", "FAQ", "cross-links to Cowork"]
+      features: ["badges (version, templates, lines)", "quiz (257 questions)", "FAQ", "cross-links to Cowork"]
       sync_with: "Guide Code (version, templates count, guide lines)"
     cowork_landing:
       local: "/Users/florianbruniaux/Sites/perso/claude-cowork-guide-landing/"
@@ -869,7 +869,7 @@ ecosystem:
       - "Cross-links modified → Update all 4 repos"
     history:
       - date: "2026-01-20"
-        event: "Code Landing sync v3.20.3, 66 templates, cross-links"
+        event: "Code Landing sync v3.20.4, 66 templates, cross-links"
         commit: "5b5ce62"
       - date: "2026-01-20"
         event: "Cowork Landing fix (paths, README, UI badges)"
diff --git a/quiz/questions/03-memory-settings.yaml b/quiz/questions/03-memory-settings.yaml
index e77063f..4e0468e 100644
--- a/quiz/questions/03-memory-settings.yaml
+++ b/quiz/questions/03-memory-settings.yaml
@@ -291,3 +291,48 @@ questions:
       file: "guide/ultimate-guide.md"
       section: "3.3 Settings & Permissions"
       anchor: "#settingsjson-team-configuration"
+
+  - id: "03-018"
+    difficulty: "senior"
+    profiles: ["senior", "power"]
+    question: "How many verification domains does the guide recommend for comprehensive quality checks?"
+    options:
+      a: "3 domains (lint, test, build)"
+      b: "8 domains: frontend, backend, types, style, performance, accessibility, security, UX"
+      c: "5 domains: syntax, logic, performance, security, style"
+      d: "12 domains covering every possible check"
+    correct: "b"
+    explanation: |
+      Boris Cherny's verification loops recommend 8 domains for comprehensive quality:
+
+      1. Frontend (UI renders correctly)
+      2. Backend (API responds correctly)
+      3. Types (TypeScript/type checks pass)
+      4. Style (linting passes)
+      5. Performance (no regressions)
+      6. Accessibility (WCAG compliance)
+      7. Security (no vulnerabilities)
+      8. UX (user flows work end-to-end)
+
+      CLAUDE.md should define which domains to check for each type of change.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "Verification Loops"
+      anchor: "#verification-loops"
+
+  - id: "03-019"
+    difficulty: "intermediate"
+    profiles: ["junior", "senior", "power"]
+    question: "What is the 'Fresh Context Pattern' and when should you use it?"
+    options:
+      a: "Clearing browser cache before testing"
+      b: "Starting a new session with clean context after implementation phase to avoid accumulated noise"
+      c: "Resetting git to a clean state"
+      d: "Deleting the .claude folder to start fresh"
+    correct: "b"
+    explanation: |
+      The Fresh Context Pattern means starting a new Claude Code session between exploration and implementation phases. During exploration, context accumulates noise (dead ends, rejected approaches, debugging output). A fresh session gives Claude clean context focused purely on implementation, improving output quality.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "Fresh Context Pattern"
+      anchor: "#fresh-context-pattern"
diff --git a/quiz/questions/04-agents.yaml b/quiz/questions/04-agents.yaml
index 897ce31..ab2d16b 100644
--- a/quiz/questions/04-agents.yaml
+++ b/quiz/questions/04-agents.yaml
@@ -291,3 +291,20 @@ questions:
       file: "guide/ultimate-guide.md"
       section: "4.4 Best Practices"
       anchor: "#explicit-activation-triggers"
+
+  - id: "04-018"
+    difficulty: "power"
+    profiles: ["power"]
+    question: "What is the invocation reliability difference between AGENTS.md (eager context) and Skills (lazy invocation)?"
+    options:
+      a: "Skills are more reliable than AGENTS.md"
+      b: "Both have 100% reliability"
+      c: "AGENTS.md: 100% reliable (always loaded) vs Skills: 53-79% auto-invoked"
+      d: "AGENTS.md is deprecated in favor of Skills"
+    correct: "c"
+    explanation: |
+      AGENTS.md uses eager context loading - always present in Claude's context, so 100% reliable activation. Skills use lazy invocation - auto-detected and invoked only when Claude recognizes the need, achieving 53-79% auto-invocation rate. Trade-off: AGENTS.md costs context tokens on every session, Skills save tokens but may miss activations. Use AGENTS.md for critical behaviors, Skills for optional capabilities.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "AGENTS.md vs Skills Evaluation"
+      anchor: "#agentsmd-vs-skills-evaluation"
diff --git a/quiz/questions/07-hooks.yaml b/quiz/questions/07-hooks.yaml
index 82f4dc4..cd423c3 100644
--- a/quiz/questions/07-hooks.yaml
+++ b/quiz/questions/07-hooks.yaml
@@ -380,3 +380,43 @@ questions:
       file: "guide/ultimate-guide.md"
       section: "7.2 Creating Hooks"
       anchor: "#hook-registration-settingsjson"
+
+  - id: "07-015"
+    difficulty: "senior"
+    profiles: ["senior", "power"]
+    question: "What configuration parameter makes a hook run asynchronously (v2.1.0+)?"
+    options:
+      a: "background: true"
+      b: "async: true"
+      c: "nonblocking: true"
+      d: "mode: async"
+    correct: "b"
+    explanation: |
+      Since Claude Code v2.1.0, hooks support `async: true` in their configuration. This makes the hook fire without blocking Claude's execution. Useful for logging, notifications, or analytics hooks where you don't need to wait for the result before continuing.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "Hook Execution Model"
+      anchor: "#hook-execution-model"
+
+  - id: "07-016"
+    difficulty: "power"
+    profiles: ["power"]
+    question: "What are the 3 limitations of async hooks compared to sync hooks?"
+    options:
+      a: "Can't access stdin, can't write files, can't read environment variables"
+      b: "No exit code feedback (can't block), no additionalContext returned, no blocking capability"
+      c: "Can't use network, can't spawn processes, can't read files"
+      d: "Limited to 1s timeout, no stderr capture, no argument passing"
+    correct: "b"
+    explanation: |
+      Async hooks trade control for speed:
+
+      1. **No exit code feedback** - Can't block Claude based on success/failure
+      2. **No additionalContext** - Can't inject context back into the conversation
+      3. **No blocking capability** - Fire-and-forget only
+
+      Use async for non-critical operations (logging, telemetry). Use sync for security gates, formatting, and validation.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "Hook Execution Model"
+      anchor: "#hook-execution-model"
diff --git a/quiz/questions/08-mcp-servers.yaml b/quiz/questions/08-mcp-servers.yaml
index 106f39a..294b166 100644
--- a/quiz/questions/08-mcp-servers.yaml
+++ b/quiz/questions/08-mcp-servers.yaml
@@ -414,3 +414,59 @@ questions:
       file: "guide/ultimate-guide.md"
       section: "4.3 Figma MCP Integration"
       anchor: "#setup-options"
+
+  - id: "08-018"
+    difficulty: "senior"
+    profiles: ["senior", "power"]
+    question: "What are the 2 core primitives of MCP Apps architecture?"
+    options:
+      a: "Functions and Routes"
+      b: "Commands and Events"
+      c: "Tools with UI metadata and UI resources (ui:// scheme)"
+      d: "Plugins and Themes"
+    correct: "c"
+    explanation: |
+      MCP Apps (SEP-1865) introduce 2 core primitives:
+
+      1. **Tools with UI metadata** - Standard MCP tools annotated with rendering hints (input forms, output displays)
+      2. **UI resources (ui:// scheme)** - Resources that return UI components instead of data
+
+      Together, these allow MCP servers to provide both functionality and user interface, enabling richer agent-to-user interactions.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "MCP Apps"
+      anchor: "#mcp-apps"
+
+  - id: "08-019"
+    difficulty: "senior"
+    profiles: ["senior", "power"]
+    question: "What does the 'auto:N' threshold control in MCP tool configuration?"
+    options:
+      a: "Maximum concurrent MCP connections"
+      b: "Cache size for tool results"
+      c: "Lazy loading - tools loaded only when description matches user intent, N = max tools to auto-load"
+      d: "Retry count for failed tool calls"
+    correct: "c"
+    explanation: |
+      The `auto:N` threshold enables lazy loading for MCP tools. Instead of loading all tools at session start (which wastes context), tools are only loaded when their description matches the user's intent. N controls the maximum number of tools to auto-load per intent match. This reduces context usage for MCP servers with many tools.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "MCP Configuration"
+      anchor: "#mcp-configuration"
+
+  - id: "08-020"
+    difficulty: "intermediate"
+    profiles: ["junior", "senior", "power"]
+    question: "Which MCP server scored highest (9.0/10) in the production ecosystem evaluation?"
+    options:
+      a: "Playwright MCP (8.8/10)"
+      b: "Context7 MCP (8.5/10)"
+      c: "Semgrep MCP - SAST, secrets, and supply chain analysis"
+      d: "Kubernetes MCP (8.4/10)"
+    correct: "c"
+    explanation: |
+      Semgrep MCP scored 9.0/10 in the production MCP ecosystem evaluation. It provides static application security testing (SAST), secrets detection, and supply chain analysis. Top scores: Semgrep (9.0), Playwright (8.8), Context7 (8.5), Kubernetes (8.4). Evaluation criteria: reliability, token efficiency, production readiness.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "MCP Servers Ecosystem"
+      anchor: "#mcp-servers-ecosystem"
diff --git a/quiz/questions/09-advanced-patterns.yaml b/quiz/questions/09-advanced-patterns.yaml
index c3d2c6a..cf00430 100644
--- a/quiz/questions/09-advanced-patterns.yaml
+++ b/quiz/questions/09-advanced-patterns.yaml
@@ -523,3 +523,173 @@ questions:
       file: "guide/methodologies.md"
       section: "Tier 4: Feature Delivery"
       anchor: "#tier-4-feature-delivery"
+
+  - id: "09-022"
+    difficulty: "power"
+    profiles: ["power"]
+    question: "What are the 5 layers of 'Mechanic Stacking'?"
+    options:
+      a: "Plan Mode, Extended Thinking, Rev the Engine, Split-Role, Permutation"
+      b: "Plan Mode, Sequential Thinking, Context7, Serena, Playwright"
+      c: "CLAUDE.md, Plan Mode, Extended Thinking, MCP Servers, Hooks"
+      d: "Plan Mode, Extended Thinking, Rev the Engine, Multi-Agent, Hooks"
+    correct: "a"
+    explanation: |
+      Mechanic Stacking layers 5 techniques for maximum Claude Code power:
+
+      1. **Plan Mode** - Safe exploration without changes
+      2. **Extended Thinking** - Deep internal reasoning
+      3. **Rev the Engine** - Progressive warm-up prompts
+      4. **Split-Role** - Separate architect/implementer/reviewer
+      5. **Permutation** - Systematic variation testing
+
+      Each layer compounds the previous one's effectiveness.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "Mechanic Stacking"
+      anchor: "#mechanic-stacking"
+
+  - id: "09-023"
+    difficulty: "power"
+    profiles: ["power"]
+    question: "What is a 'Permutation Framework' in Claude Code?"
+    options:
+      a: "A/B testing of UI variants"
+      b: "CLAUDE.md-driven systematic variation testing: define dimensions, generate variants, implement, evaluate"
+      c: "Random code generation for benchmarks"
+      d: "Automated regression test suite"
+    correct: "b"
+    explanation: |
+      A Permutation Framework uses CLAUDE.md to drive systematic variation testing. The workflow: define variation dimensions (e.g., algorithm choice, data structure, caching strategy), generate all combinations, implement each variant, evaluate with metrics. This ensures you explore the solution space methodically rather than picking the first approach.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "Permutation Frameworks"
+      anchor: "#permutation-frameworks"
+
+  - id: "09-024"
+    difficulty: "senior"
+    profiles: ["senior", "power"]
+    question: "What mental model describes the developer as an orchestrator of Claude instances?"
+    options:
+      a: "Pipeline Manager"
+      b: "Agent Supervisor"
+      c: "You Are the Main Thread - CPU scheduler analogy where developer dispatches tasks to agents"
+      d: "Task Router"
+    correct: "c"
+    explanation: |
+      "You Are the Main Thread" uses the CPU scheduler analogy: the developer is the main thread dispatching work to Claude instances (worker threads). You manage priorities, context switches, and synchronization. Key insight: you don't write code - you manage the agents that write code. This shifts the skill from coding to orchestration.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "You Are the Main Thread"
+      anchor: "#you-are-the-main-thread"
+
+  - id: "09-025"
+    difficulty: "senior"
+    profiles: ["senior", "power"]
+    question: "When task list items consistently diverge from actual work done, what does the guide recommend?"
+    options:
+      a: "Delete the task list and start over"
+      b: "Add more granular detail to each task"
+      c: "Use divergence patterns as diagnostic: too broad = break down, too narrow = step back, wrong priorities = re-align with goals"
+      d: "Ignore the divergence and continue"
+    correct: "c"
+    explanation: |
+      Task list divergence is diagnostic, not failure. Patterns reveal issues:
+
+      - **Too broad** tasks → break down into smaller pieces
+      - **Too narrow** tasks → step back and think bigger
+      - **Wrong priorities** → re-align tasks with actual goals
+      - **Consistent drift** → your mental model of the problem is wrong
+
+      Use divergence as signal, not noise.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "Task Management Patterns"
+      anchor: "#task-management-patterns"
+
+  - id: "09-026"
+    difficulty: "senior"
+    profiles: ["senior", "power"]
+    question: "What is the 'occurrence rule' for claiming established patterns in code reviews?"
+    options:
+      a: ">5 occurrences = established, 2-5 = emerging, <2 = not established"
+      b: ">10 occurrences = established, 3-10 = emerging, <3 = not established"
+      c: "Based on file count, not occurrences"
+      d: ">20 occurrences = established, 5-20 = emerging, <5 = not established"
+    correct: "b"
+    explanation: |
+      Anti-hallucination safeguards for code reviews define the occurrence rule:
+
+      - **>10 occurrences** = established pattern (safe to reference)
+      - **3-10 occurrences** = emerging pattern (mention cautiously)
+      - **<3 occurrences** = NOT an established pattern (don't claim it is)
+
+      This prevents Claude from hallucinating "established conventions" from a few examples.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "Anti-Hallucination Safeguards"
+      anchor: "#anti-hallucination-safeguards"
+
+  - id: "09-027"
+    difficulty: "power"
+    profiles: ["power"]
+    question: "What are the 3 specialized agents in multi-agent code review?"
+    options:
+      a: "Security Reviewer, Performance Analyst, UX Reviewer"
+      b: "Linter, Type Checker, Test Runner"
+      c: "Consistency Auditor, SOLID Analyst, Defensive Code Auditor"
+      d: "Junior Reviewer, Senior Reviewer, Architect Reviewer"
+    correct: "c"
+    explanation: |
+      Multi-agent PR review uses 3 specialized agents:
+
+      1. **Consistency Auditor** - Checks naming conventions, import patterns, code style adherence
+      2. **SOLID Analyst** - Reviews architectural principles, dependency injection, single responsibility
+      3. **Defensive Code Auditor** - Validates error handling, input validation, edge cases
+
+      Each agent has anti-hallucination safeguards (occurrence rule, file-scoped claims).
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "Multi-Agent PR Review"
+      anchor: "#multi-agent-pr-review"
+
+  - id: "09-028"
+    difficulty: "senior"
+    profiles: ["senior", "power"]
+    question: "What is 'comprehension debt' according to Addy Osmani's '80% Problem'?"
+    options:
+      a: "Documentation debt - missing docs"
+      b: "Code you shipped but don't fully understand - distinct from technical debt"
+      c: "Review debt - unreviewed PRs"
+      d: "Design debt - skipped design phase"
+    correct: "b"
+    explanation: |
+      Comprehension debt is code you shipped but don't fully understand. It's distinct from technical debt (code you understand but know is suboptimal). With AI-generated code, comprehension debt grows silently: you accept suggestions without deep understanding. The fix: always be able to explain WHY, not just WHAT.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "Practitioner Insights"
+      anchor: "#practitioner-insights"
+
+  - id: "09-029"
+    difficulty: "senior"
+    profiles: ["senior", "power"]
+    question: "What is the '4-step cycle' for CLAUDE.md as compounding memory (Boris Cherny)?"
+    options:
+      a: "Write, Test, Deploy, Monitor"
+      b: "Plan, Execute, Review, Commit"
+      c: "Error, Rule, Read, Never repeated"
+      d: "Ask, Implement, Validate, Ship"
+    correct: "c"
+    explanation: |
+      Boris Cherny's mental model for CLAUDE.md as compounding memory:
+
+      1. **Error** - Claude makes a mistake
+      2. **Rule** - You add a rule to CLAUDE.md preventing it
+      3. **Read** - Claude reads CLAUDE.md on every session start
+      4. **Never repeated** - The mistake never happens again
+
+      Goal: never correct Claude twice for the same mistake. CLAUDE.md compounds over time.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "Boris Cherny Mental Models"
+      anchor: "#boris-cherny-mental-models"
diff --git a/quiz/questions/10-reference.yaml b/quiz/questions/10-reference.yaml
index 5406e05..e638ceb 100644
--- a/quiz/questions/10-reference.yaml
+++ b/quiz/questions/10-reference.yaml
@@ -463,3 +463,54 @@ questions:
       file: "guide/ultimate-guide.md"
       section: "10.2 Keyboard Shortcuts"
       anchor: "#input--navigation"
+
+  - id: "10-019"
+    difficulty: "junior"
+    profiles: ["junior", "senior", "power"]
+    question: "What does the '--from-pr' flag do in Claude Code v2.1.27?"
+    options:
+      a: "Creates a new pull request"
+      b: "Resumes sessions linked to a GitHub PR number or URL"
+      c: "Fetches all PR comments into context"
+      d: "Closes a pull request"
+    correct: "b"
+    explanation: |
+      The `--from-pr` flag (v2.1.27) lets you resume Claude Code sessions linked to a specific GitHub PR. Usage: `claude --from-pr 123` or `claude --from-pr https://github.com/org/repo/pull/123`. This loads the PR context (diff, comments, checks) directly into the session.
+    doc_reference:
+      file: "guide/claude-code-releases.md"
+      section: "v2.1.27"
+      anchor: "#v2127"
+
+  - id: "10-020"
+    difficulty: "senior"
+    profiles: ["senior", "power"]
+    question: "What breaking change was introduced in Claude Code v2.1.19 for hook/command arguments?"
+    options:
+      a: "Arguments removed entirely"
+      b: "Migration from $ARGUMENTS.0 (dot syntax) to $ARGUMENTS[0] (bracket syntax)"
+      c: "New $PARAMS variable replaced $ARGUMENTS"
+      d: "Arguments now require JSON format"
+    correct: "b"
+    explanation: |
+      Claude Code v2.1.19 changed argument access syntax from dot notation (`$ARGUMENTS.0`) to bracket notation (`$ARGUMENTS[0]`). This breaking change affects all custom commands and hooks that reference arguments. Update your scripts to use the new bracket syntax.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "Breaking Changes"
+      anchor: "#breaking-changes"
+
+  - id: "10-021"
+    difficulty: "junior"
+    profiles: ["junior", "senior", "power", "pm"]
+    question: "According to the guide's 'Myths vs Reality' section, what is the truth about 'hidden Claude Code features'?"
+    options:
+      a: "Secret flags exist for power users"
+      b: "Hidden features require an enterprise plan"
+      c: "There are no hidden features - Anthropic uses progressive rollout, not secret flags"
+      d: "Beta flags are stored in .claude/config"
+    correct: "c"
+    explanation: |
+      The Myths vs Reality section debunks the "hidden features" myth. Anthropic uses progressive rollout (features gradually enabled for users) rather than secret flags or hidden configurations. All features are documented in release notes. There is no secret power-user mode.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "Myths vs Reality"
+      anchor: "#myths-vs-reality"
diff --git a/quiz/questions/11-learning-with-ai.yaml b/quiz/questions/11-learning-with-ai.yaml
index 764b0a6..fbd0810 100644
--- a/quiz/questions/11-learning-with-ai.yaml
+++ b/quiz/questions/11-learning-with-ai.yaml
@@ -257,3 +257,49 @@ questions:
       file: "guide/learning-with-ai.md"
       section: "Embracing AI Tools"
       anchor: "#embracing-ai-tools"
+
+  - id: "11-016"
+    difficulty: "senior"
+    profiles: ["senior", "power"]
+    question: "What are the 3 failure modes identified in Addy Osmani's '80% Problem' of agentic coding?"
+    options:
+      a: "Context loss, token overflow, hallucination"
+      b: "Overengineering, assumption propagation, sycophantic agreement"
+      c: "Slow responses, wrong language, missed tests"
+      d: "Memory leaks, race conditions, deadlocks"
+    correct: "b"
+    explanation: |
+      Addy Osmani's "80% Problem" identifies 3 failure modes:
+
+      1. **Overengineering** - AI adds unnecessary complexity (abstraction layers, premature optimization)
+      2. **Assumption propagation** - One wrong assumption cascades through the codebase
+      3. **Sycophantic agreement** - AI agrees with bad ideas instead of pushing back
+
+      The 80% refers to AI getting you 80% of the way, with the last 20% requiring deep human understanding.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "Practitioner Insights"
+      anchor: "#practitioner-insights"
+
+  - id: "11-017"
+    difficulty: "intermediate"
+    profiles: ["junior", "senior", "power"]
+    question: "What are the 3 symptoms of vibe coding 'context overload'?"
+    options:
+      a: "Memory leaks, slow startup, API rate limits"
+      b: "Big-bang context dumps, 5K+ line prompts, performance degradation"
+      c: "Lost files, broken imports, missing tests"
+      d: "Excessive commits, merge conflicts, branch proliferation"
+    correct: "b"
+    explanation: |
+      Context overload symptoms in vibe coding:
+
+      1. **Big-bang context dumps** - Pasting entire codebases into prompts
+      2. **5K+ line prompts** - Exceeding effective context window usage
+      3. **Performance degradation** - Claude's quality drops as context grows
+
+      The fix: progressive disclosure (feed context incrementally), use CLAUDE.md for persistent context, and start fresh sessions between phases.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "Vibe Coding Anti-Patterns"
+      anchor: "#vibe-coding-anti-patterns"
diff --git a/quiz/questions/12-architecture.yaml b/quiz/questions/12-architecture.yaml
index e67f337..80bd786 100644
--- a/quiz/questions/12-architecture.yaml
+++ b/quiz/questions/12-architecture.yaml
@@ -206,3 +206,54 @@ questions:
       file: "guide/architecture.md"
       section: "Sub-Agent Types"
       anchor: "#sub-agent-types"
+
+  - id: "12-013"
+    difficulty: "junior"
+    profiles: ["junior", "senior", "power"]
+    question: "What tool replaced TodoWrite for task management in Claude Code v2.1.16+?"
+    options:
+      a: "ProjectManager API"
+      b: "Kanban tool"
+      c: "Tasks API (TaskCreate, TaskGet, TaskList, TaskUpdate)"
+      d: "AgendaWrite tool"
+    correct: "c"
+    explanation: |
+      The Tasks API (v2.1.16+) replaced TodoWrite with 4 tools: TaskCreate (create tasks with descriptions), TaskGet (fetch full task details), TaskList (list all tasks with status), TaskUpdate (update status, dependencies, metadata). Key improvement: task dependencies via blockedBy/blocks fields.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "Tasks API"
+      anchor: "#tasks-api"
+
+  - id: "12-014"
+    difficulty: "power"
+    profiles: ["power"]
+    question: "What is the API cost overhead of fetching full details for 50 tasks via Tasks API?"
+    options:
+      a: "10x overhead (10 API calls)"
+      b: "25x overhead (25 API calls)"
+      c: "51x overhead (1 TaskList + 50 TaskGet calls required)"
+      d: "100x overhead (100 API calls)"
+    correct: "c"
+    explanation: |
+      Tasks API limitation: TaskList returns only summary fields (id, subject, status, owner, blockedBy). To get full details (description, metadata), you need individual TaskGet calls. For 50 tasks: 1 TaskList + 50 TaskGet = 51 API calls. This N+1 pattern is a known limitation compared to TodoWrite which returned everything in one call.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "Tasks API Limitations"
+      anchor: "#tasks-api-limitations"
+
+  - id: "12-015"
+    difficulty: "power"
+    profiles: ["power"]
+    question: "What is TeammateTool and what is its status?"
+    options:
+      a: "Production-ready multi-agent framework"
+      b: "Experimental multi-agent coordination (spawnTeam, discoverTeams, requestJoin, approveJoin) - unstable, no official support"
+      c: "Official Anthropic tool for team collaboration"
+      d: "Deprecated feature replaced by Tasks API"
+    correct: "b"
+    explanation: |
+      TeammateTool is an experimental multi-agent coordination tool with 4 operations: spawnTeam, discoverTeams, requestJoin, approveJoin. Status: unstable, not officially supported by Anthropic. It allows Claude instances to form teams and coordinate, but the API is subject to change without notice. Not recommended for production use.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "TeammateTool"
+      anchor: "#teammatetool"
diff --git a/quiz/questions/13-security.yaml b/quiz/questions/13-security.yaml
index 9b34ba0..ab631b2 100644
--- a/quiz/questions/13-security.yaml
+++ b/quiz/questions/13-security.yaml
@@ -172,3 +172,37 @@ questions:
       file: "guide/security-hardening.md"
       section: "Security Posture Levels"
       anchor: "#security-posture-levels"
+
+  - id: "13-011"
+    difficulty: "senior"
+    profiles: ["senior", "power"]
+    question: "Which sandbox isolation approach combines microVM isolation with network policies for agent autonomy?"
+    options:
+      a: "E2B (hosted cloud sandboxes)"
+      b: "Fly.io Sprites (edge compute)"
+      c: "Docker Sandboxes (with custom templates and network policies)"
+      d: "Cloudflare Sandbox SDK"
+    correct: "c"
+    explanation: |
+      Docker Sandboxes provide microVM-level isolation with customizable network policies. Key features: custom Dockerfile templates for reproducible environments, network policies to control egress/ingress, volume mounts for persistent storage, and CPU/memory limits. This approach suits teams wanting full control over sandbox configuration while maintaining strong isolation.
+    doc_reference:
+      file: "guide/sandbox-isolation.md"
+      section: "Docker Sandboxes"
+      anchor: "#docker-sandboxes"
+
+  - id: "13-012"
+    difficulty: "junior"
+    profiles: ["junior", "senior", "power", "pm"]
+    question: "What is the GitHub Issue Auto-Creation Bug (#13797) and why is it dangerous?"
+    options:
+      a: "Issues get automatically deleted"
+      b: "Claude Code accidentally creates public GitHub issues containing private project details"
+      c: "Issues are created but remain private"
+      d: "Only affects paid enterprise accounts"
+    correct: "b"
+    explanation: |
+      The GitHub Issue Auto-Creation Bug (#13797) causes Claude Code to accidentally create public GitHub issues containing private project details. Over 17 accidental disclosures documented, affecting v2.0.65+. The danger: internal code, architecture decisions, and private context leak publicly. Mitigation: disable the gh CLI tool or restrict GitHub permissions.
+    doc_reference:
+      file: "guide/known-issues.md"
+      section: "GitHub Issue Auto-Creation Bug"
+      anchor: "#github-issue-auto-creation-bug"
diff --git a/quiz/questions/14-privacy-observability.yaml b/quiz/questions/14-privacy-observability.yaml
index 5cd435b..af66b3e 100644
--- a/quiz/questions/14-privacy-observability.yaml
+++ b/quiz/questions/14-privacy-observability.yaml
@@ -172,3 +172,25 @@ questions:
       file: "guide/observability.md"
       section: "Native Commands"
       anchor: "#native-commands"
+
+  - id: "14-011"
+    difficulty: "intermediate"
+    profiles: ["junior", "senior", "power"]
+    question: "When should you use 'Co-Authored-By' vs 'Assisted-By' in git commits?"
+    options:
+      a: "Always use Co-Authored-By for any AI involvement"
+      b: "Co-Authored-By when AI generated significant code, Assisted-By when AI only advised or reviewed"
+      c: "Never attribute AI in commits"
+      d: "Use both on every commit"
+    correct: "b"
+    explanation: |
+      AI traceability in git commits:
+
+      - **Co-Authored-By**: When AI generated significant portions of code (implementation, refactoring)
+      - **Assisted-By**: When AI only advised, reviewed, or suggested minor changes
+
+      This distinction matters for code ownership, audit trails, and understanding who (or what) actually wrote the code. Some teams add this to CLAUDE.md as a mandatory commit convention.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "AI Traceability"
+      anchor: "#ai-traceability"
diff --git a/quiz/questions/15-ai-ecosystem.yaml b/quiz/questions/15-ai-ecosystem.yaml
index 943059f..19463e3 100644
--- a/quiz/questions/15-ai-ecosystem.yaml
+++ b/quiz/questions/15-ai-ecosystem.yaml
@@ -279,3 +279,66 @@ questions:
       file: "guide/ai-ecosystem.md"
       section: "8. Cost & Subscription Strategy"
       anchor: "#8-cost--subscription-strategy"
+
+  - id: "15-011"
+    difficulty: "senior"
+    profiles: ["senior", "power"]
+    question: "What cost savings does the Claude Code to LM Studio bridge provide?"
+    options:
+      a: "20-30% savings on API costs"
+      b: "50% savings but slower execution"
+      c: "80-90% savings by planning with Opus then executing free with local LLM"
+      d: "No savings but better privacy"
+    correct: "c"
+    explanation: |
+      The Claude Code to LM Studio bridge provides 80-90% cost savings. Strategy: use Opus for planning phase (~$0.50-2 per session) then execute implementation with a free local LLM via LM Studio. The bridge script translates Claude's plan into local LLM instructions. Trade-off: local LLMs are less capable but free.
+    doc_reference:
+      file: "examples/scripts/bridge.py"
+      section: "LM Studio Bridge"
+      anchor: "#lm-studio-bridge"
+
+  - id: "15-012"
+    difficulty: "power"
+    profiles: ["power"]
+    question: "What are the 3 documented external orchestration systems for multi-agent Claude Code?"
+    options:
+      a: "LangChain, AutoGPT, CrewAI"
+      b: "Vercel AI SDK, Fly.io Machines, Railway"
+      c: "Gas Town (Steve Yegge), multiclaude (dlorenc), agent-chat (Justin Abrahms)"
+      d: "Kubernetes, Docker Compose, Terraform"
+    correct: "c"
+    explanation: |
+      Three documented external orchestration systems:
+
+      1. **Gas Town** (Steve Yegge) - Multi-agent coordination with shared context
+      2. **multiclaude** (dlorenc) - Parallel Claude Code instances with task distribution
+      3. **agent-chat** (Justin Abrahms) - Inter-agent communication protocol
+
+      These are community-built tools, not official Anthropic products. Each solves multi-agent coordination differently.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "External Orchestration"
+      anchor: "#external-orchestration"
+
+  - id: "15-013"
+    difficulty: "senior"
+    profiles: ["senior", "power"]
+    question: "What 3 areas should sub-agents audit when evaluating a skeleton project before forking?"
+    options:
+      a: "Performance, Cost, Scalability"
+      b: "UI, API, Database"
+      c: "Security, Architecture, Developer Experience (DX)"
+      d: "Tests, Documentation, CI/CD"
+    correct: "c"
+    explanation: |
+      Skeleton project evaluation uses 3 specialized sub-agents:
+
+      1. **Security** - Audit dependencies, secrets management, auth patterns, known vulnerabilities
+      2. **Architecture** - Evaluate code organization, patterns, scalability, maintainability
+      3. **Developer Experience (DX)** - Check setup complexity, documentation quality, tooling, onboarding friction
+
+      Each agent produces a score and recommendations before the fork decision.
+    doc_reference:
+      file: "guide/ultimate-guide.md"
+      section: "Skeleton Projects Workflow"
+      anchor: "#skeleton-projects-workflow"