Curate a weekly newsletter from RSS feeds
A weekly industry newsletter typically takes four hours to produce. Task: A weekly industry newsletter typically takes four hours to produce. Two hours reading sources, one hour picking what matters, one hour writing intros. An OpenClaw content discovery agent can cut that to under 30 minutes. It monitors 120 RSS feeds, scores every article against configurable relevance criteria, compiles a ranked digest, sends a Telegram preview each Thursday morning, and publishes to ConvertKit with one tap. Total human time per issue: about 18 minutes. The agent handles the discovery; the editor handles the judgment. This guide walks through every piece of that pipeline: the skills, the cron jobs, the scoring logic, the Telegram preview loop, and the email platform connection. If you have a working OpenClaw installation (start with our setup guide if you do not), you can build this in an afternoon. How the Newsletter Curation Pipeline Works Most newsletter automation guides describe a flat pipe: fetch feeds, summarize, email. That misses the step that matters most: deciding what deserves your readers’ attention. An RSS reader gives you everything. A curation agent gives you the ten items worth reading, ranked and annotated. Here is the weekly cadence for a typical setup: Continuous (every 6 hours): OpenClaw monitors RSS feeds, Reddit, X, and Hacker News for new content matching your topic scope. New items land in a workspace digest file with metadata. Wednesday night (cron, 11 PM): A scoring skill runs across the week’s collected items, ranking each by relevance, recency, source authority, and engagement signals. The top 10-15 items are pulled into a draft digest. Thursday morning (heartbeat, 8 AM): OpenClaw compiles the digest with editorial intros, formats it for your newsletter template, and sends a Telegram preview with approve/revise options. Your review (15-20 minutes): You scan the digest, reorder if needed, tweak an intro, and tap approve. Post-approval (automated): OpenClaw pushes the final newsletter to your email platform via API, logs what was covered in memory, and archives the issue. Each step is a separate skill. The cron jobs and heartbeat wire them together. Let us build each one. Setting Up Multi-Source Content Monitoring RSS feeds are the backbone, but newsletters that only pull from RSS miss the conversations happening on Reddit, X, and Hacker News. The content monitoring skill watches all four. The Content Monitor Skill Create skills/newsletter-monitor.md in your OpenClaw workspace: Configuring Your Sources Create workspace/newsletter/sources.md: The source authority tiers feed into the scoring skill later. Tier 1 content gets a scoring boost; Tier 3 content needs stronger engagement signals to make the cut. Scheduling the Monitor Wire the monitor to run every 6 hours: The --session isolated flag keeps each monitoring run in its own context, preventing inbox bloat from accumulating in a long-running session. For more on session management, see our cron jobs guide. After a week of monitoring, you will have 80-200 items in the inbox depending on how many sources you track. The next skill separates signal from noise. Scoring and Filtering Content This is the step that separates a newsletter agent from an RSS reader. The scoring skill evaluates every inbox item against four criteria and produces a ranked shortlist. The Scoring Skill Create skills/newsletter-score.md: Why Configurable Weights Matter It is worth testing several scoring configurations. Early versions that weight recency too heavily tend to fill newsletters with breaking news that ages poorly by the time readers open the email. Shifting weight toward relevance and source authority produced digests that readers forwarded more often. You should tune these weights for your audience. A daily tech digest needs higher recency weight. A monthly strategy newsletter needs higher relevance and authority weight. The point is that the weights are visible in the skill file, not buried in a black box. Running the Scorer Schedule it for Wednesday night, giving the week’s content time to accumulate: Compiling the Digest with Editorial Intros Raw links with summaries are not a newsletter. Readers subscribe for the editor’s perspective on why each item matters. The compilation skill writes those editorial intros. The Digest Compiler Skill Create skills/newsletter-compile.md:; tools: OpenClaw. Human role: The original post does not say how a person stayed involved.; limitations: Requires a working OpenClaw installation and API keys for email platforms. Human approval is mandatory via Telegram to prevent publishing errors. Costs vary by model usage; digest compilation is the most expensive stage. Source reliability is secondary, as claims are not independently verified..
Tools
Published on Agent Work Atlas:
Overview
A weekly industry newsletter typically takes four hours to produce.
Workflow
A weekly industry newsletter typically takes four hours to produce. Two hours reading sources, one hour picking what matters, one hour writing intros. An OpenClaw content discovery agent can cut that to under 30 minutes. It monitors 120 RSS feeds, scores every article against configurable relevance criteria, compiles a ranked digest, sends a Telegram preview each Thursday morning, and publishes to ConvertKit with one tap. Total human time per issue: about 18 minutes. The agent handles the discovery; the editor handles the judgment. This guide walks through every piece of that pipeline: the skills, the cron jobs, the scoring logic, the Telegram preview loop, and the email platform connection. If you have a working OpenClaw installation (start with our setup guide if you do not), you can build this in an afternoon. How the Newsletter Curation Pipeline Works Most newsletter automation guides describe a flat pipe: fetch feeds, summarize, email. That misses the step that matters most: deciding what deserves your readers’ attention. An RSS reader gives you everything. A curation agent gives you the ten items worth reading, ranked and annotated. Here is the weekly cadence for a typical setup: Continuous (every 6 hours): OpenClaw monitors RSS feeds, Reddit, X, and Hacker News for new content matching your topic scope. New items land in a workspace digest file with metadata. Wednesday night (cron, 11 PM): A scoring skill runs across the week’s collected items, ranking each by relevance, recency, source authority, and engagement signals. The top 10-15 items are pulled into a draft digest. Thursday morning (heartbeat, 8 AM): OpenClaw compiles the digest with editorial intros, formats it for your newsletter template, and sends a Telegram preview with approve/revise options. Your review (15-20 minutes): You scan the digest, reorder if needed, tweak an intro, and tap approve. Post-approval (automated): OpenClaw pushes the final newsletter to your email platform via API, logs what was covered in memory, and archives the issue. Each step is a separate skill. The cron jobs and heartbeat wire them together. Let us build each one. Setting Up Multi-Source Content Monitoring RSS feeds are the backbone, but newsletters that only pull from RSS miss the conversations happening on Reddit, X, and Hacker News. The content monitoring skill watches all four. The Content Monitor Skill Create skills/newsletter-monitor.md in your OpenClaw workspace: Configuring Your Sources Create workspace/newsletter/sources.md: The source authority tiers feed into the scoring skill later. Tier 1 content gets a scoring boost; Tier 3 content needs stronger engagement signals to make the cut. Scheduling the Monitor Wire the monitor to run every 6 hours: The --session isolated flag keeps each monitoring run in its own context, preventing inbox bloat from accumulating in a long-running session. For more on session management, see our cron jobs guide. After a week of monitoring, you will have 80-200 items in the inbox depending on how many sources you track. The next skill separates signal from noise. Scoring and Filtering Content This is the step that separates a newsletter agent from an RSS reader. The scoring skill evaluates every inbox item against four criteria and produces a ranked shortlist. The Scoring Skill Create skills/newsletter-score.md: Why Configurable Weights Matter It is worth testing several scoring configurations. Early versions that weight recency too heavily tend to fill newsletters with breaking news that ages poorly by the time readers open the email. Shifting weight toward relevance and source authority produced digests that readers forwarded more often. You should tune these weights for your audience. A daily tech digest needs higher recency weight. A monthly strategy newsletter needs higher relevance and authority weight. The point is that the weights are visible in the skill file, not buried in a black box. Running the Scorer Schedule it for Wednesday night, giving the week’s content time to accumulate: Compiling the Digest with Editorial Intros Raw links with summaries are not a newsletter. Readers subscribe for the editor’s perspective on why each item matters. The compilation skill writes those editorial intros. The Digest Compiler Skill Create skills/newsletter-compile.md:
Reported result
The source claims this reduces production time from 4 hours to under 30 minutes, with 15-20 minutes of human review. It estimates API costs at $10-15/month. However, no independent metrics or verified outcomes are provided to confirm these savings or cost figures.
Human role
The original post does not say how a person stayed involved.
Limitations
Requires a working OpenClaw installation and API keys for email platforms. Human approval is mandatory via Telegram to prevent publishing errors. Costs vary by model usage; digest compilation is the most expensive stage. Source reliability is secondary, as claims are not independently verified.
Try it
I want to set up an OpenClaw newsletter curation pipeline for [replaceable: industry topic]. First, create a workspace with sources.md listing [replaceable: 5-10 RSS feeds and subreddits]. Configure a monitor skill to run every 6 hours, saving new items to an inbox. Next, set up a scoring skill to rank items by relevance and recency, selecting the top 10. Then, create a compilation skill that drafts editorial intros based on [replaceable: voice guidelines]. Finally, configure a Telegram preview step that sends the draft for my approval before publishing to [replaceable: email platform]. Run a dry test with 3 sample items first. Do not publish to real subscribers until I confirm the preview format and content quality. If any step fails, stop and report the error.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.