Knowledge Capture

Knowledge Capture

This skill enables the persistent storage of valuable insights discovered during a chat session. It extracts, structures, and saves knowledge into a local knowledge/ folder.

Workflow

1. Identify the Signal: When the user requests to "capture" or "save" knowledge, identify the most relevant high-signal information from the recent conversation history.
2. Select a Template: Review TEMPLATES.md and choose the best fit for the content (Research Summary, Technical Snippet, Industrial Trends, or Challenges).
3. Format the Content: Structure the raw information using the chosen template. Ensure technical snippets are correctly highlighted and summaries are concise but comprehensive.
4. Determine Topic and Type: Define a short, descriptive topic (e.g., "transformer-scaling") and the content type (e.g., "Research Summary").
5. Execute Save: Use the scripts/save_knowledge.py script to save the file.

Commands

Run the saving script with the following structure:

python3 knowledge-capture/scripts/save_knowledge.py \
  --topic "<topic-name>" \
  --type "<template-type>" \
  --content "<formatted-markdown-content>" \
  --root "."

Guidelines

• Surgical Extraction: Don't just dump the entire chat history. Focus on the core value: the "lesson learned" or the "final solution."
• Consistent Naming: The script handles naming (YYYY-MM-DD-topic.md), so ensure the topic provided is descriptive and URL-safe (the script will normalize it).
• Quality: Ensure the Markdown is clean and readable. Use code blocks for technical snippets.