Q&A Appender

Core Responsibilities

Usage Scenarios

/qa

/ask

• Have specific doubts (What/Why/How) about a certain point in an existing document
• Received answers through conversations
• Need to record the Q&A in notes

Core Logic: Knowledge Classification

• Context Heavy: Unintelligible without the original context → Append to the end of the original text
• Atomic Concept: Universal concept that can be understood independently of the original text → Generate an independent knowledge card

references/classification_guide.md

Workflow

/qa

/ask

1. Intelligent Session Range Recognition
   • Automatic Recognition: Analyze recent conversations to identify the current discussion topic and question boundaries
   • Key Signals:
     • Time point when the user raised a new question
     • Topic transition markers (e.g., "Additionally", "Another question")
     • Problem resolution markers (e.g., "Okay", "Got it", "Resolved")
   • Extraction Principle: Only extract continuous conversation segments relevant to the current question, avoid including irrelevant content
   • Intelligent Judgment:
     • If it's a single Q&A (1-3 rounds): Extract the Q&A
     • If it's a complex discussion (4-10 rounds): Extract the complete discussion process
     • If it exceeds 10 rounds: Prompt the user to confirm whether to include all content
2. Read Classification Criteria
   • Must read

     references/classification_guide.md

     before execution
   • Determine the attribution of the Q&A (Context Heavy vs Atomic Concept)
3. Determine Processing Method
   • Context Heavy: Append the

     ## 💡 Practical Q&A

     section at the end of the original text
   • Atomic Concept: Generate an independent knowledge card
4. Apply C+I Structure
   • Read

     references/ci_structure.md

     to obtain complete specifications
   • Strictly organize content according to the template format
   • Logic (Insight) comes first, Context comes below
5. Obtain Confirmation
   • Provide the following information at once:
     • Session Range: Clearly inform which rounds of conversations were extracted (e.g., "Extracted the last 3 rounds of conversations")
     • Judgment Result: Context Heavy or Atomic Concept
     • Judgment Reason: Why this judgment was made
     • Target File Path: Which file will be written to
   • Prohibited: Displaying a complete preview of the content to be appended (the preview is too long and interferes with reading)
   • Confirm only once and wait for the user's reply
6. Write to File
   • Directly write to the file after user confirmation
   • Prohibited: Displaying the preview or confirming again
   • Briefly inform the result after writing

Execution Specifications (Important)

• ✅ Automatically identify relevant conversation segments and avoid extracting irrelevant content
• ✅ Clearly inform the extracted session range during the confirmation phase
• ❌ Do not extract the entire conversation history

• ✅ Confirm only once in step 5
• ❌ Prohibited to confirm again in step 6

• ✅ Only display the session range, judgment result, and brief reason
• ❌ Prohibited to display the complete preview of the content to be appended
• Reason: Long preview content interferes with reading, and users trust the judgment logic

Step 1: Intelligent session range recognition (internal processing)
Step 2-4: Read + analyze + organize content (internal processing)
Step 5: Output session range + judgment result + wait for confirmation (once)
Step 6: Directly write + brief notification (completed)

C+I Structure Overview

• C (Context) - Original Context: The complete conversation at that time, including code, error messages, and parameters
• I (Insight) - Key Takeaway: A one-sentence universal principle without specific cases

references/ci_structure.md

Collaboration with Other Skills

• conversation-extractor

  : Generates basic notes

• process-doc-generator

  : Generates process documents

• The appended document can be further expanded using

  process-doc-generator

Resource Files

• C+I Structure Details: references/ci_structure.md
• Classification Guide: references/classification_guide.md