briefed_skill

Overview

This skill sets up and runs Briefed, a personal newsletter intelligence pipeline that fetches Gmail newsletters, summarizes articles with Claude Haiku, and serves a local web reader with voting, notes, and interest tracking. It provides a daily digest flow: pre-fetch Gmail metadata, generate AI summaries, fetch full HTML bodies, run a local Express reader, and send a notification when the digest is ready. The stack uses Python scripts for Gmail handling, Claude Haiku for summarization, and a Node.js web app for reading and interaction.

How this skill works

A pre-fetch script pulls compact metadata from your Gmail account into newsletter-inbox.json and filters transactional mail. A scheduled Claude Haiku agent reads that compact JSON and writes AI summaries into newsletter-today.json, scoring items by interest weights. A fetch-bodies script attaches full HTML bodies, and an Express server exposes a local reader with APIs for voting, notes, and saving. A cron/agent sequence sends a notification when the digest is ready so you can open the reader.

When to use it

• You want a daily digest that reduces newsletter reading time with concise AI summaries.
• You need a local, private newsletter reader with voting, notes, and interest-based ranking.
• You want to aggregate and prioritize email newsletters from a Gmail account.
• You need automated inbox intelligence that filters transactional mail.
• You want to run summaries with Claude Haiku and host a local web app for interaction.

Best practices

• Install gogcli and authenticate a read-only Gmail OAuth account before running scripts.
• Tune newsletter-interests.json topic weights to match your preferences for better ranking.
• Adjust SKIP_SUBJECT_PATTERNS and SKIP_SENDERS in pre-fetch.py to reduce transactional noise.
• Run the pipeline via a daily cron or LaunchAgent so summaries appear automatically each morning.
• Keep the reader behind a local network or VPN if exposing the port externally; use notification channels for alerts only.

Example use cases

• Morning briefing: get 7am summaries of the most relevant newsletters scored by your interests.
• Research triage: surface and save links into a reading list for deeper follow-up.
• Team share: run Briefed on a collective account and broadcast the digest to a Slack or Discord channel.
• Inbox hygiene: automatically filter out transactional newsletters and focus on curated articles.
• Note-taking workflow: attach per-story notes and votes to build a searchable archive of important items.