jovaneyck/obsidian-ai-agent-stuff/commands

Generate Knowledge Vault Quiz

You are going to create and run an interactive quiz based on the contents of this knowledge vault using ChromaDB.

Quiz Parameters

Number of Questions: {{num_questions}} (default: 10 if not specified)

Topic Focus: {{topic}} (if specified, focus all questions on this specific topic; if not specified, make the quiz as broad and diverse as possible across all vault content)

Instructions

1. Topic Discovery Using ChromaDB:

• Use ChromaDB MCP tools to access the "notes" collection
• If a specific topic is provided:
• Use mcp__chroma__chroma_query_documents to search for documents semantically related to the topic
• Limit to 10-15 most relevant documents for focused quiz content
• If no topic is specified:
• First, get the total document count using mcp__chroma__chroma_get_collection_count
• Generate random offset values dynamically using Python or command-line tools
• Use mcp__chroma__chroma_get_documents with these random offsets to get diverse content
• Aim for 15-20 diverse documents across different domains

1. Document Retrieval and Content Analysis:

• After identifying relevant documents from ChromaDB, use the Read tool to access the actual markdown files
• Extract file paths from the ChromaDB metadata and read only those specific files
• Analyze content for key concepts, frameworks, and interconnected ideas
• Focus on understanding, not just facts

1. Question Generation:

• If a specific topic is provided, generate ALL questions about that topic only
• If no topic is specified, create the most diverse quiz possible - pull from different domains (software craftsmanship, cognitive science, systems thinking, AI, leadership, coaching, etc.)
• Use only content from this knowledge vault - don't inject external knowledge
• Create multiple choice questions with 4 options each
• Ensure questions test understanding, not just memorization. You can do this by providing out-of-order listings, paraphrasing etc.
• CRITICAL: When creating multiple choice options, ensure that each option is meaningfully different and tests real understanding. Avoid creating options that are merely reordered lists of the same elements.

1. Quiz Execution:

• Act as an interactive quizmaster
• Present questions one at a time
• Wait for user response after each question
• Provide immediate feedback (correct/incorrect)
• If incorrect, briefly explain the mistake and the correct answer
• Track score throughout

1. Quiz Completion:

• Show final score at the end
• Provide constructive, positive feedback
• Give an experience level assessment based on performance
• CRITICAL: Provide context-specific next steps based on the quiz topic and user's performance:
• For software craftsmanship topics: suggest specific practices, books, or techniques to try
• For cognitive science topics: recommend specific exercises or areas to explore
• For systems thinking topics: suggest frameworks or real-world applications
• For coaching topics: recommend specific coaching techniques or situations to practice
• Always make suggestions actionable and specific to the quiz content

Example Usage

/generate-quiz num_questions:5 topic:Conway's Law

(Creates 5 questions focused only on Conway's Law using semantic search)

/generate-quiz num_questions:15

(Creates 15 diverse questions across all vault topics using random sampling)

/generate-quiz

(Creates 10 diverse questions across all vault topics using random sampling)

Important Notes

• Questions should reflect the interconnected nature of concepts in the vault
• Use the same positive, constructive tone as established in previous interactions
• Ensure questions are challenging but fair
• Draw from notes, literature notes, and project content
• Reference specific files when explaining answers if relevant
• Efficiency: Use ChromaDB to efficiently discover relevant content without reading all files

Random Offset Generation Strategy

For diverse content sampling, use this approach to generate smart random offsets:

1. Calculate Safe Maximum: max_offset = total_documents - limit (where limit is your per-query limit, e.g., 5)
2. Generate Multiple Offsets: Create 6-8 different random offsets to sample across the collection
3. Ensure Distribution: Aim for offsets that span the full range (early, middle, and late documents)
4. Example Calculation:

• If collection has 267 documents and using limit=3
• max_offset = 267 - 3 = 264
• Generate offsets like: 12, 67, 145, 198, 234, 89

Command Examples:

# Python method (most reliable)
python -c "import random; offsets=[random.randint(0, 264) for _ in range(6)]; print(offsets)"
 
# PowerShell method (Windows)
$offsets = 1..6 | ForEach-Object { Get-Random -Maximum 264 }; Write-Output $offsets

Quiz Mode Selection

After generating the quiz content, ALWAYS ask the user to choose between:

A) Interactive Quiz Mode: Run the quiz directly in the conversation (traditional Claude quizmaster experience) B) Add to Web App: Add the quiz to the quiz web app in scripts/quiz-app/quiz-engine.js

Wait for the user's selection before proceeding with either option.

Implementation Steps

1. Check ChromaDB collection info and document count using mcp__chroma__chroma_get_collection_info and mcp__chroma__chroma_get_collection_count
2. If topic specified:

• Use mcp__chroma__chroma_query_documents with the topic as query text
• Limit results to 10-15 most semantically similar documents

1. If no topic specified:

• Generate random offset values using one of these methods:
• Python: python -c "import random; print(' '.join(str(random.randint(0, {max_offset})) for _ in range(6)))"
• Where {max_offset} = total_documents - 5 (to ensure we don't exceed bounds with limit)
• Use mcp__chroma__chroma_get_documents with these random offsets to sample diverse content
• Use small limits (3-5 docs each) to ensure we stay within collection bounds

1. Extract file paths from ChromaDB metadata and use Read tool to access actual content
2. Generate quiz questions with context-specific performance feedback for different score ranges
3. MANDATORY: Present the quiz mode selection to the user and wait for their choice
4. Based on user selection:

• Option A: Run the interactive quiz as quizmaster with context-specific conclusions
• Option B: Add the quiz to the web app including performance feedback data structure

Begin generating quiz content now, but remember to ask about quiz mode before execution.