Skill: Conversation Scribe

Description

This skill enables the agent to take intelligent, structured notes on conversations. It filters out noise, categorizes content into Projects (Resolutions, Ideas, Logs), and automatically indexes them into a project-centric knowledge base.

Usage

Use this skill when the user says "take a note", "summarize this conversation", "save this solution", or asks to record the current interaction.

Core Logic: The Intelligent Filter

Before writing a note, you must process the conversation history using the following steps:

1. Noise Reduction:

   • Identify the final correct answer or agreed-upon solution.
   • Discard failed attempts, "I apologize" messages, or "Did you mean X?" clarification loops.
   • Extract specific context definitions (e.g., "NCU means National Central University").

2. Project Classification:

   • Determine the Project Name based on context (e.g., "Gemini_CLI", "Website", "Research").
   • If no project is clear, use 00_Inbox.

3. Category Selection:

   • Resolutions: A specific problem was solved, a bug fixed, or a definition established. (High Value)
   • Include
     • Workflow
     • Program Structure
     • Think Process
     • And many more...
   • Ideas: Brainstorming, future plans, or architectural designs.
   • Logs: General status updates, "what I did today", or chronological records.

Execution

Use the run_shell_command tool to execute the python script.

# Template
run_shell_command(f"python3 {{skill_path}}/scripts/write_note.py --project '{{Project_Name}}' --category '{{Category}}' --title '{{Short_Title}}' --summary '{{One_Sentence_Summary}}' --content '{{Full_Markdown_Content}}'")

Content Formatting Rules

• Summary: A single, clean sentence describing the note's value.
• Content: Use Markdown.
  • For Resolutions, use headers like ### Context, ### Problem, ### Solution.
  • For Ideas, use bullet points.
  • Ensure all newlines in the content string are escaped or passed correctly (the script handles \n replacement).

Examples

Scenario 1: Resolving a definition confusion

User: "What is NCU?" -> Agent: "North Central?" -> User: "No, National Central U." -> Agent: "Ok." -> User: "Note this."

summary = "Established definition of NCU as National Central University."
content = "### Context\n- **NCU**: National Central University.\n\n### Resolution\n- Model previously hallucinated. Added definition to project context."
run_shell_command(f"python3 {{skill_path}}/scripts/write_note.py --project 'Research_Project' --category 'Resolutions' --title 'NCU_Definition' --summary '{summary}' --content '{content}'")

Scenario 2: Quick idea for Inbox

User: "Remind me to buy milk and fix the roof."

run_shell_command(f"python3 {{skill_path}}/scripts/write_note.py --project '00_Inbox' --title 'Chores_List' --summary 'Personal to-do list.' --content '- Buy milk\n- Fix roof'")

Critical Instruction on Context

Always check if a _project_summary.md exists in the project folder before answering deep questions. It contains the "State" of the project.