curation
Skill99 starsChanged 6 months agoApache-2.0View on GitHub
Selection and ranking rules — how to turn the raw fetched pool into the handful of items the reader actually wants, guided by the taste profile in interests.json.
In this repository27 other files it gives its agents.
- ai-radio/AGENTS.mdAGENTS.md
- customer-support/AGENTS.mdAGENTS.md
- data-analyst/AGENTS.mdAGENTS.md
- document-processor/AGENTS.mdAGENTS.md
- news-curator/AGENTS.mdAGENTS.md
- repo-maintainer/AGENTS.mdAGENTS.md
- seo-auditor/AGENTS.mdAGENTS.md
- audio-mixingSkill
- cover-image-generationSkill
- metadata-generationSkill
- music-generationSkill
- researchSkill
- script-writingSkill
- tts-generationSkill
- memorySkill
- scannerSkill
- data-explorerSkill
- python-dataSkill
- pdf-parsingSkill
- reconciliationSkill
- slide-creatorSkill
- vendor-verificationSkill
- briefingSkill
- deliverSkill
- fetch-newsSkill
- github-researcherSkill
- git-workflowSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 24
Collapse duplicates of one story: When several outlets carry the same development, pick the strongest single link and mention coverage …
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Curation Skill
Curation is editing, not summarizing. From the candidate pool in
new_items.json, choose the items this specific reader should see today.
Selection Rules
- The taste profile decides. Re-read
interests.jsonbefore selecting.preferences.moreentries boost an item;preferences.lessentries drop it unless it is genuinely major;preferences.notescarry dated, specific guidance from past feedback — treat recent notes as the strongest signal. - Cap the briefing at 10 news items total across all topics, and up to 4 community-buzz items. Fewer is fine; padding is not. If a topic produced nothing worth including, say so in one line rather than including filler.
- Prefer, in order: primary sources and official announcements → original
reporting → high-engagement community discussion → aggregation/commentary.
Drop listicles and rumor pieces by default (they are on the default
lesslist). - Collapse duplicates of one story. When several outlets carry the same development, pick the strongest single link and mention coverage breadth in the why-it-matters line ("also covered by Reuters and The Verge") — breadth itself is a signal the story matters.
- Genre feeds are the serendipity channel. Pick at most 2-3 genre items the reader did not explicitly ask about but would plausibly care about, given their profile. Label them as discoveries.
- Community items are for signal, not noise. Select a Hacker News item only when it adds something the news coverage lacks: practitioner experience, technical pushback, or early buzz on something not yet covered.
- Longform earns its slot with depth. Select at most 2 Medium pieces per briefing, and only for genuine substance — tutorials, architecture write-ups, first-hand experience reports. Medium has no engagement signal in the feed, so judge entirely on title substance and taste-profile fit; skip anything that reads as engagement-bait or paywalled teaser fluff.
Ranking Within the Briefing
Lead with the item that best combines: relevance to an explicit topic, recency, source quality, and taste-profile fit. Genre discoveries and community buzz always come after the reader's explicit topics.
What Never Gets Selected
- Items you cannot link.
- Items whose headline alone is too thin to say why it matters — do not pad with invented detail.
- Pure engagement-bait framed as news, regardless of topic match.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

