knowledge-brain_skill

Overview

This skill captures, summarizes, and organizes knowledge from URLs, YouTube videos, documents, and pasted content into a structured personal knowledge base. It creates raw extracts and concise summaries, maintains a lightweight index in agent memory, and proactively recalls stored knowledge when relevant to a user query. The skill enforces consistent file, date, and summary conventions to keep the collection searchable and durable.

How this skill works

On input the skill detects the source type (web page, YouTube, local file, audio, or pasted text), extracts the full raw content, and writes a raw markdown file with standardized frontmatter. It then generates a 100–400 word summary into a paired summary file and places both into either a matching category or an unsorted folder. After changes it updates a compact index in agent memory, appends entries to a changelog, and optionally initializes git for version control. When answering questions the skill consults the index, searches summaries, loads relevant items, synthesizes across entries, and cites source titles and IDs.

When to use it

• Save an article, video, or file you want preserved and summarized.
• Ask a question where past captures might be relevant — the skill will proactively search the knowledge base.
• Bulk-import an existing folder of notes or a collection of files into the structured KB.
• Request sorting or reorganization of unsorted entries into clear categories.
• Rename, split, or merge categories while preserving history and updating the index.

Best practices

• Choose a stable knowledge base path on first run and let the skill store it in memory for future use.
• Name files using the prescribed slug--YYYY-MM-DD convention to keep IDs consistent and traceable.
• Prefer summaries of 100–400 words and include a short Key Points list for fast recall.
• Keep category _category.md files concise so automated proposals can match unsorted entries reliably.
• Commit meaningful operations with one commit per logical change when using git; do not push without explicit user instruction.

Example use cases

• Capture a long investigative article: extract the body, save a raw file, and generate a concise summary with takeaways.
• Add a YouTube lecture: download and clean transcript, insert time markers, and create paired raw/summary files.
• Sort a backlog of unsorted captures: propose category assignments, confirm, and move files while updating the changelog.
• Answer a design question using previously captured architecture notes: consult index, load summaries, synthesize recommendations, and cite sources.
• Import a colleague’s folder of PDFs and markdown notes into the KB with an initial changelog and index update.