English | 简体中文

Capture Learning Notes

A portable Codex Skill for collecting articles, blog posts, and complete podcast episodes into a durable Markdown-and-HTML knowledge base.

This public export intentionally excludes video and Xiaohongshu capture while preserving the original onboarding, vault setup, URL deduplication, topic indexes, daily digests, HTML reading views, and podcast transcription workflow.

Capabilities

• Initialize and reuse one canonical knowledge-base vault across conversations.
• Capture complete webpages and blog articles with evidence-aware summaries.
• Capture complete podcast episodes from public transcripts or local Whisper transcription.
• Store podcast transcript assets in Markdown, plain text, and JSONL formats.
• Deduplicate source URLs and maintain library, topic, podcast, transcript, and daily indexes.
• Generate linked HTML reading views while keeping Markdown as the source of truth.
• Preserve first-run onboarding, user ratings, and user-authored reflections.

Installation

Clone this repository into your personal Codex Skills directory:

git clone https://github.com/Unmoored-Lab/capture-learning-notes.git ~/.codex/skills/capture-learning-notes

The Skill becomes available in a new Codex conversation.

Example usage

Use $capture-learning-notes to collect this podcast episode into my knowledge base: <URL>

You can also provide a webpage, a small batch of links, or a link with one sentence describing what to focus on.

Repository contents

• SKILL.md: core workflow, safety rules, and completion contract.
• agents/openai.yaml: Codex UI metadata.
• assets/: starter vault, HTML renderer, and podcast note template.
• references/: onboarding, knowledge-base format, note schema, and podcast workflow.
• scripts/: vault setup, URL extraction, deduplication, indexing, rendering setup, and local podcast transcription.

Optional local dependencies

Podcast transcription is installed only when first needed and requires ffmpeg, ffprobe, and a supported local Whisper backend. HTML rendering uses markdown-it-py in a vault-local environment.

License

MIT © 2026 Jared Chen