K12 Smart Teacher

In K12 education tutoring, students often require personalized explanations and practice when encountering incorrect answers, but traditional approaches face a set of specific problems. Parents or teachers struggle to quickly pinpoint a student's weak knowledge points, the explanation process can be dry and abstract, generated practice questions lack targeting, and learning progress is difficult to track systematically. This skill aims to solve these pain points by providing a structured intelligent tutoring framework.

Specific Problems Addressed

• Inefficient Error Tutoring: Correcting and analyzing errors on student homework or test papers is time-consuming and prone to missing key misconceptions when done manually.
• Abstract Explanation Methods: Pure text or verbal explanations struggle to help students intuitively understand complex concepts, especially abstract topics like mathematical formulas or physical principles.
• Unsystematic Practice Generation: Manually creating questions makes it hard to ensure difficulty gradation, often leading to inconsistent question quality or misalignment with grade levels.
• Fragmented Learning Records: Students' error histories and knowledge point mastery are scattered, lacking mechanisms for continuous tracking and periodic reviews.

How the Skill Works

This skill is a complete K12 intelligent tutoring system. Its core capabilities and key workflow steps are designed around the above issues:

Core Capabilities and Workflow

1. Intelligent Recognition and Grading: The system automatically analyzes images of student homework or exams uploaded by the user, identifying the subject and grade level (e.g., determining if it's middle school physics or elementary math based on content), then grades each question and records correct and incorrect answers.

2. Visual Error Explanation: Explanation is the core step, strictly following the principle of "explain first, then practice". The fixed workflow includes:

   • Introducing the knowledge point with a relatable, life-based analogy.
   • Using graphics, comics, or flowcharts for visual demonstration of the problem-solving process.
   • Explaining the reasoning in steps and contrasting incorrect vs. correct understanding.
   • Mandatory use of the web_search tool to recommend 1-2 quality solution videos from platforms like Bilibili or Xueersi to ensure explanation completeness.

3. Graduated Practice Generation: Based on the knowledge points from errors, the system reads topics from resources like references/balanced_topic_bank.json and generates practice sets containing three difficulty tiers: Basic Consolidation (⭐), Ability Enhancement (⭐⭐), and Expansion Challenge (⭐⭐⭐), with a maximum of 8 questions per set, prioritizing quality over quantity.

4. Learning Archive and Progress Tracking: All interaction information—including basic student details, error records, knowledge point mastery levels, and recommended videos—is updated into a MEMORY.md file. The system can intelligently recommend review topics based on historical records, enabling continuous tracking.

5. Output and Fault Tolerance: Practice files are generated in Word format (relying on python-docx) by default. If this fails, the system automatically degrades to Markdown and informs the user clearly in Chinese. The process avoids exposing raw English error stacks.

Scope and Limitations

This skill is not omnipotent; its capability boundaries must be noted during use:

• Dependency on External Tools: Feature implementation relies on the quality of image OCR recognition, internet search capabilities (for video recommendation), and libraries like python-docx. If tools are unavailable, the system degrades gracefully but does not interrupt the core tutoring flow.
• Not for Non-Academic Content: It focuses on tutoring for nine core subjects (Chinese, Math, English, Physics, Chemistry, Biology, History, Geography, Politics) and does not handle non-academic content such as medical advice, psychological diagnoses, or admission policies.
• Limited Textbook Precision: If the user does not provide textbook versions, chapters, or original question materials, the system generates generic exercises. Users can supply materials for the skill to perform secondary refinement.
• Input Quality Requirements: Recognition may fail if images are blurry or questions are obscured, requiring the user to resend clear images or input questions manually. The system provides friendly prompts instead of error exits.