prashantkr0923-commits

LinkedIn Newsletter Digest Agent — Testing skill for Claude Code

Testing community

Agent that turns a week of LinkedIn newsletter emails into one ranked Sunday digest and learns from three one-tap marks: Worth it, Read later, Not interested.

How to install LinkedIn Newsletter Digest Agent

This entry records only its repository, not the path inside it, so there is no exact command to give. Open prashantkr0923-commits/linkedin-newsletter-digest-agent and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What LinkedIn Newsletter Digest Agent does

Agent that turns a week of LinkedIn newsletter emails into one ranked Sunday digest and learns from three one-tap marks: Worth it, Read later, Not interested. In testing, 70% of its top 10 were articles the reader liked, against 44% for newest first. Built on Google Apps Script and the Claude API. In development, open-source release planned.

Alternatives in Testing

Baro — A CLI that turns a goal into a pull request - and a sandbox for testing concurrent AI coding agents on the Moz 120 ★
Claude Dev Skill — A Claude Code custom skill that turns Claude into a Tech Lead — coordinating parallel Worker Agents through a 112 ★
Spec-Flow Agents — Spec-Flow coordinates multiple coding tools through a single shared canon 73 ★

README

LinkedIn Newsletter Digest Agent

An agent that turns a week of LinkedIn newsletter emails into one ranked digest every Sunday morning, and learns what to put on top from three one-tap marks: **Worth it**, **Read later** and **Not interested**.

**Status: in development.** Version 1 is being built and tested on my own inbox. It will be released as open source for anyone to run in their own Google account once it has run live for four Sundays. The source code is in a private repository until then.

The problem

Subscribing to LinkedIn newsletters is easy; keeping up with them isn't. A busy subscription list sends dozens of editions a week as separate emails. Most go unread, and the few worth reading are buried among near-daily ones that never are.

Email filters can't tell a good edition from a bad one, and LinkedIn's own feed ranks by what is popular, not by what a particular reader finds worth their time.

How it works

flowchart LR
    A[LinkedIn newsletter emails in Gmail] --> B[Fetch: read-only]
    B --> C[Parse: newsletter, author, title, link, text]
    C --> D[Enrich: summary, topic, format, advert flag]
    D --> E[Rank: newsletter and topic track records]
    E --> F[Sunday email: ranked digest]
    F --> G[One-tap marks]
    G --> E

Fetch. Every Sunday, the agent reads the week's newsletter emails through the Gmail API, with read-only access.
Parse. Each email becomes one edition: the newsletter, author, title, a clean article link and the article text. The email is cut at its footer before anything is read, so unsubscribe links and their personal tokens are never touched.
Enrich. A language model writes a 2 to 4 sentence summary and labels the topic, the format, the angle and whether the edition is an advert or a job list. Every reply is checked against the allowed labels before it is stored.
Rank. Each edition is scored from the reader's own track record with its newsletter (60%) and its topic (40