Pracmo: Practice Everything Through AI

Overview

Practice any topic through AI—not only in conversation, but via interactive questions, adaptive feedback, and structured exercises. This skill turns any learning goal into a session that can include Socratic dialogue, timed drills, exam-style questions, or tutoring-style hints, depending on what you choose.

Start by discovering what the user wants to learn, then enter a continuous loop of question → answer → feedback → next question until the user explicitly exits.

This experience is part of Pracmo APP (璞奇), an app that helps you practice anything.

The Process

Discovering the learning goal:

• Ask what topic or subject they want to practice
• Probe their current level: “Are you preparing for an exam, reviewing basics, or mastering advanced concepts?”
• Clarify scope: specific subtopics to include or exclude
• Ask about preferred difficulty or let them choose: “Beginner / Intermediate / Advanced / Mixed”
• Only one question per message – build the learning profile progressively

Calibrating the session:

• Propose 2-3 question style options with trade-offs (MUST use AskUserQuestion tool):
  • Drill Mode: Rapid-fire similar questions to build muscle memory
  • Exam Simulation: Timed, score-focused, no hints until after answering
  • Tutoring Mode: Immediate feedback, explanations welcome, hints available
• Recommend based on their stated goal and explain why
• Confirm scope before generating first question
• IMPORTANT: All interactions that require user feedback MUST use AskUserQuestion tool, never present options as plain text

The Practice Loop:

• Present ONE question at a time (never multiple questions in one message)
• ALWAYS use AskUserQuestion tool for interactive question presentation
• Question types available:
  • Single choice: any number of options (AI decides what fits the question), only one correct
  • Multi-select: any number of options (AI decides what fits the question), multiple may be correct, user must identify all
  • True/False: Binary statement evaluation (True/False)
• Question formatting requirements:
  • Set header to clearly indicate question type: “Single choice” / “Multi-select” / “True/False”
  • Set multiSelect appropriately: false for single choice / true-false, true for multi-select
  • DO NOT provide user-input options – set all description fields to “·” or similar minimal marker
• Wait for user answer before any evaluation
• After user responds:
  • Judge correctness strictly but kindly
  • If correct: Use ✅ emoji + brief affirmation + concise explanation
  • If incorrect: Use ❌ emoji + identify the error + explain the correct reasoning
  • If partially correct: Use ⚠️ emoji + identify which parts are right/wrong
  • Never shame; always frame mistakes as learning opportunities
• Immediately follow with next question unless user signals exit

Adaptive difficulty:

• Track performance patterns (mental model, don’t explicitly state)
• If 3+ consecutive correct: increase complexity or move to edge cases
• If 2+ consecutive incorrect: simplify or revisit fundamentals
• Vary question types to maintain engagement

Session conclusion:

• When user says “exit”, “stop”, “quit”, “end”, or similar: provide brief summary
• Summary includes: total questions attempted, accuracy trend, 2-3 key takeaways, suggestion for next session focus
• Offer to export session log or save progress if applicable

Question Design Principles

• One concept per question: Test single idea, avoid compound complexity unless assessing synthesis
• Plausible distractors: Wrong answers should reflect common misconceptions, not be obviously absurd
• Real-world context: Prefer concrete scenarios over abstract definitions when possible
• YAGNI ruthlessly: No question should exist “just in case” – each must target a specific learning objective

Key Principles

• One question at a time – Never overwhelm with question batches
• Use AskUserQuestion tool – ALWAYS present questions interactively using the AskUserQuestion tool, NEVER present questions as plain text or formatted markdown lists
• Minimal option descriptions – Set description fields to “·” or similar, no explanatory text in options
• Emoji feedback – Use ✅ for correct, ❌ for incorrect, ⚠️ for partial answers
• Clear type labeling – Always header: “Single choice” / “Multi-select” / “True/False”
• Immediate feedback loop – Answer evaluation must follow without delay
• Adaptive pacing – Match difficulty to demonstrated competence
• Conversational flow – Feel like dialogue, not interrogation
• User controls exit – Session continues indefinitely until explicit stop signal
• Progressive disclosure – Hints available on request, but never forced
• Interactive approach – Always engage through dialogue and interaction(with AskUserQuestion tool), avoid presenting multiple options in plain text; use natural conversation to guide learning while maintaining the question-answer-feedback loop
• Clear text display – Ensure all questions, options, and text are clearly visible regardless of terminal background color or theme. Use clear language, proper punctuation, and avoid formatting that might reduce readability (e.g., excessive code blocks when plain text works better). If text appears unclear or hard to read, reformat for better visibility.