Claude Code Guide: 271 Questions (2026)

The claude-code-ultimate-guide by Florian Bruniaux (4K+ stars) is the most exhaustive Claude Code reference available. At 22K+ lines, it covers every feature with 271 quiz questions, 225 templates, 41 Mermaid diagrams, and a threat model containing 655 known malicious skill definitions.

What It Is

A monolithic reference guide that attempts to document every aspect of Claude Code. It combines tutorial content, reference material, assessment tools, and security guidance into a single repository.

The structure:

  • Core Documentation — 22K+ lines covering installation through advanced patterns
  • Quiz Bank — 271 questions with answers for self-assessment and team certification
  • Template Library — 225 ready-to-use configurations (CLAUDE.md rules, hooks, commands, MCP configs)
  • Mermaid Diagrams — 41 visual flowcharts explaining workflows and architectures
  • Threat Model — 655 documented malicious or deceptive skill definitions with detection rules

Why It Matters

The threat model is the standout feature. As the Claude Code skill ecosystem grows, so does the attack surface. Malicious skills can exfiltrate code, inject backdoors, or modify files in ways that look legitimate. The 655-entry database documents known attack patterns and provides detection rules.

The quiz bank is the only assessment tool available for Claude Code proficiency. Teams use it for onboarding validation — new developers work through the questions to prove they understand Claude Code’s behavior before getting unsupervised access.

The template library overlaps with claude-code-templates but focuses on security-conscious configurations that account for the threat model.

Installation

Clone for local access:

git clone https://github.com/FlorianBruniaux/claude-code-ultimate-guide.git ~/references/claude-code-ultimate-guide

Browse the Quiz

cd ~/references/claude-code-ultimate-guide
open quizzes/index.md

Access the Threat Model

open threat-model/malicious-skills-db.json

Use Templates

ls templates/
# claude-md/
# hooks/
# commands/
# mcp-configs/
# settings/

Copy templates into your project:

cp templates/claude-md/secure-default.md /path/to/project/CLAUDE.md

Key Features

  1. 271 Quiz Questions — multiple choice and scenario-based questions covering basic usage through advanced security topics. Self-grading with explanations.

  2. 225 Templates — CLAUDE.md files, hook scripts, command definitions, MCP configurations, and settings files. Each template is documented with use cases and customization notes.

  3. 655 Malicious Skills DB — documented attack patterns including code exfiltration, prompt injection via skills, file modification attacks, and credential theft. Each entry includes detection signatures.

  4. 41 Mermaid Diagrams — visual representations of Claude Code’s architecture, session lifecycle, tool execution flow, MCP communication, and security boundaries.

  5. 22K+ Lines of Content — this isn’t a lightweight guide. Every feature gets thorough treatment with examples, edge cases, and gotchas.

  6. Security-First Templates — configurations that enforce file access restrictions, command allowlists, and output validation.

  7. Team Certification Path — a structured sequence of quizzes that teams use to certify Claude Code proficiency levels (beginner, intermediate, advanced, security).

  8. Detection Rules — regex and heuristic rules for scanning skill files for known malicious patterns. Integrate into CI/CD to catch dangerous skills before installation.

Real Usage Example

Running a Self-Assessment

## Quiz: CLAUDE.md Basics (15 questions)
Q1: Where does Claude Code look for CLAUDE.md files?
a) Only the project root
b) Project root and ~/.claude/
c) Project root, ~/.claude/, and every parent directory
d) Only ~/.claude/
Answer: c — Claude Code searches the project root, ~/.claude/ (global),
and every parent directory up to the filesystem root.

Using the Threat Model

{
  "id": "MAL-042",
  "name": "Silent File Exfiltration",
  "category": "data_theft",
  "description": "Skill that instructs Claude Code to include file contents in web fetch requests disguised as API calls",
  "detection": "regex: fetch.*base64.*readFile|curl.*--data.*cat",
  "severity": "critical",
  "first_seen": "2026-01-15"
}

Scan your installed skills:

# Check skills against known malicious patterns
cd ~/references/claude-code-ultimate-guide
python scripts/scan-skills.py /path/to/project/.claude/

Applying a Secure Template

cp templates/claude-md/secure-default.md /path/to/project/CLAUDE.md

The secure default template includes:

## Security Rules
- NEVER include file contents in web requests
- NEVER execute commands that pipe to external URLs
- NEVER modify .git/hooks/ or .github/workflows/ without explicit approval
- NEVER install packages not listed in the approved dependencies
- Log all tool invocations to .claude/audit-log.txt

When To Use

  • Security-conscious teams — the threat model and detection rules are unmatched elsewhere
  • Team onboarding — the quiz bank provides structured assessment
  • Certification programs — organizations use the quizzes for internal Claude Code certification
  • Template shopping — 225 templates cover more scenarios than most other collections
  • Audit preparation — the security documentation helps justify Claude Code usage to compliance teams

When NOT To Use

  • Quick start needs — 22K lines is too much for someone who just wants to get running; use claude-howto instead
  • Casual personal projects — the security focus adds friction that solo side-projects don’t need
  • Staying current — the guide updates less frequently than rapidly evolving repos like awesome-claude-code

FAQ

How long does the full quiz take?

271 questions at ~1 minute each = ~4.5 hours. Most teams break it into the 4 certification levels and complete one per session.

Is the malicious skills database complete?

No database is complete. The 655 entries cover known patterns as of the last update. New attack vectors emerge regularly. Use it as a baseline, not a guarantee.

Can I contribute quiz questions?

Yes. The repo accepts PRs for new questions following the schema in CONTRIBUTING.md. Each question needs a correct answer, explanation, and difficulty level.

How do the detection rules work?

Regex patterns matched against skill file contents. Run the scanner script against your .claude/ directory to flag potential matches. False positives are possible — review each flag manually.

Our Take

8/10. The threat model alone justifies bookmarking this repo. No other resource catalogs malicious Claude Code skill patterns with this level of detail. The quiz bank is the only assessment tool available for the ecosystem. Loses points because the 22K-line guide is intimidating and the update frequency lags behind faster-moving repos. Best for teams that need security assurance and structured onboarding — not for individual developers looking for quick tips.

Configure MCP → Build your server config with our MCP Config Generator.

See Also

Try it: Browse 155+ skills in our Skill Finder.