News Digest
Ihre Themen, täglich gescannt, als ein gerankter Überblick geliefert
Scannt Ihre Themen täglich in RSS, Reddit, X und Hacker News und liefert einen gerankten, gut lesbaren Überblick.

In 3 Schritten installieren
Keine Anmeldung. Keine CLI-Tools. Wählen Sie Ihre AI, kopieren Sie eine Zeile und fügen Sie sie ein.
Lieber selbst machen? Kaufen Sie den Skill ($4.99), laden Sie die ZIP-Datei aus Ihrer Kaufbestätigung herunter, entpacken Sie sie in das oben angezeigte Verzeichnis und starten Sie Ihre AI-Sitzung neu — gleiches Ergebnis. Haben Sie ihn schon gekauft? Ihr AI-Assistent kann ihn für Sie herunterladen.
Beweise aus der Praxis
Echte Menschen haben genau diese Aufgabe mit AI-Agenten erledigt.
Curate a weekly newsletter from RSS feeds
Anwendungsfall ansehen →AnwendungsfallSend a daily AI news digest on a schedule
Anwendungsfall ansehen →Was ist enthalten
Eine SKILL.md (die Anleitung) + eine Startvorlage. Eine Vorschau:
News Digest
When to use
- You spend an hour a day scanning the same feeds and still feel behind
- You run a newsletter and curation eats your whole week
- Your team needs to stay current on a niche topic without hiring a researcher
- Your source list keeps growing faster than your attention
- You want "one email a day" instead of fifty open tabs
Workflow
- Define the brief (input: topics + trusted sources → output: source list)
Name your topics and your trusted sources: RSS feeds, subreddits, X searches, Hacker News. A small tight list beats 120 feeds on day one — start with your 10 best sources and expand only when the digest feels thin. Note why each source is trusted; that note is what stops source rot later. - Fetch and dedupe (input: source list + schedule → output: clean item pool)
Pull on a fixed schedule — daily for a personal digest, or every 6 hours into a store for a weekly newsletter. Then dedupe aggressively: the same story always arrives from five sources wearing different headlines. One story, one entry, the best link kept. - Score and rank (input: clean items → output: ranked shortlist)
Score every item on four dimensions — relevance, freshness, authority, engagement — and apply a duplication penalty when near-identical stories slip through. The weights are tunable and they should be: start with defaults, then adjust weekly based on what you actually clicked. Publish the rubric so you can tell when the ranking drifts. - Write the digest (input: shortlist → output: summaries + intro)
Write one short summary per item — faithful to the source, no clickbait rewrites. Add a "why it matters" line and a link to the original. For public newsletters, write an editorial intro in your voice (under 150 words) and label your own opinions as opinions. What you can't verify, you flag — never present a rumor as fact. - Preview then send (input: draft digest → output: delivered digest)
Always run one full trial before turning on the schedule: review the output, tune the prompt, fix the weights. For newsletters, the draft goes to a preview channel first (e.g. Telegram) and sends only on your explicit approval. Then deliver on schedule: email, chat app, or your notes — one fixed time, every time.
Output spec
- Source list: each source with a one-line "why trusted" note; additions and removals logged
- Scoring rubric: relevance / freshness / authority / engagement + duplication penalty, weights adjustable
- Digest format: intro (≤150 words, public digests) + ranked items, each with title / one-line summary / source link / date / why-it-matters
- Fixed schedule: daily or weekly, one send time
- Every item links to its original source — no exceptions
Good vs. bad examples
Vorschau der Startvorlage
News Digest — Brief & Edition Template
Two pages. Fill the brief once (review monthly); use the edition sheet per issue.
Brief (fill once)
Topics: {topic 1}, {topic 2}, {topic 3} Schedule: {daily at 07:00 / weekly Thursday} Delivery: {email to …, Telegram channel …, Notion page …}
Sources
Scoring rubric (weights add to 10)
Duplication penalty: {e.g. "−2 if the same story appears in 3+ sources — keep the best link"}
Edition — {date}
Trial run (first issue only)
[ ] Ran once manually [ ] Reviewed output [ ] Tuned prompt/weights [ ] Schedule turned on
Ranked shortlist
Das Skill-Paket selbst (SKILL.md + Vorlage) ist auf Englisch verfasst.
Das könnte Ihnen auch gefallen
Grok Bot Routine Rescue
Find which scheduled routine is eating your weekly pool — and kill it
Workflows
Inbox Triage
Ihr Postfach, vorsortiert: Dringendes markiert, Antworten entworfen, Sie klicken nur noch auf Senden
Workflows
Bill Fighter
Fechten Sie Überpreise und abgelehnte Ansprüche mit beleggestützten Schreiben an