new-term-radar
신규 AI 용어를 매일 추적, X·Hugging Face·GitHub 등에서 모델·제품·레포 이름을 빠르게 수집
아직 감지된 업데이트가 없습니다. GitHub 릴리스가 나오면 여기에 보입니다.
개인 계정의 공개 저장소 소유자는 GitHub로 로그인해 운영자를 확인하고 소개를 직접 관리할 수 있습니다. 관리자 승인은 필요하지 않습니다.
로그인한 뒤 확인 파일을 만들어 이 저장소의 기본 브랜치에 30분 안에 커밋해 주세요. 다른 브랜치나 포크의 파일은 인정하지 않습니다. 파일을 올린 뒤 아래에서 확인하면 관리자 승인 없이 관리할 수 있습니다. 자격을 유지하려면 파일을 남겨 두세요.
로그인이 필요하면 로그인 화면으로 이동합니다. 로그인 후 이곳으로 돌아와 원하는 확인 버튼을 다시 눌러 주세요.
프로젝트 폴더에서 /nomorevibe verify를 실행하면 소유권을 확인하고 소개를 직접 관리할 수 있습니다.
1분 만에 확인하기An Agent Skill that offers three angles for a social post, then writes the one you pick.
저장소 README에서 가져옴
README 전체를 불러오는 중입니다. 아래는 저장된 발췌입니다.
An Agent Skill that offers you three different angles for a social post, then writes the one you pick.
It stops at the draft. It never invents the personal detail. Those two rules are
the whole design, and the rest of this file explains why.
Someone told me they stay with their scheduling tool mostly because of its post
templates. When they run out of ideas, the templates give them a way back in.
That stuck with me. It is not the part of the job I would have guessed people
get stuck on.README 전문 보기 ↗An Agent Skill that offers you three different angles for a social post, then writes the one you pick.
It stops at the draft. It never invents the personal detail. Those two rules are the whole design, and the rest of this file explains why.
Someone told me they stay with their scheduling tool mostly because of its post templates. When they run out of ideas, the templates give them a way back in. That stuck with me. It is not the part of the job I would have guessed people get stuck on.
Asking an agent to "write me a post" has the same problem from the other side. You get one draft, produced on the first guess. You read it, you are not convinced, and there is nothing specific to push back on, so you rewrite it yourself and the agent saved you nothing.
This changes the first move. The skill looks at what you have already published, offers three angles from three different categories, and asks which one. Then it asks for the one thing only you have: the mistake, the number, the name. Then it writes from that.
It never invents the personal detail. Every angle asks for a specific fact that has to come from you, and the skill is told, more than once, that inventing it is the failure mode of the whole thing. A post with a fabricated failure in it reads plausible, is false, and goes out under your own name.
It stops at the draft. Nothing is published or scheduled from this skill. It shows you the finished text and hands off; publishing stays a separate, deliberate step.
skills/publora-post-ideas/SKILL.md: how the agent chooses angles, what it
has to ask before drafting, what it checks in the text, and what it is never
allowed to invent.skills/publora-post-ideas/references/angles.md: the angle library across
eight categories: Story, Behind the scenes, Tip, Case study, List, How-to,
Opinion, Question. Each entry names the specific fact that has to come from you.references/hooks.md: what the first two lines have to do before the feed
collapses the rest, and the openings that reliably fail.references/voice.md: how to read a voice off someone's existing posts rather
than asking them to describe it, and what must never be carried over from a
source article.references/machine-tells.md: the patterns that give a draft away after every
automated check has passed.references/platforms.md: how one angle changes shape across LinkedIn, X,
Threads, Bluesky, Telegram and the networks that need media first.One command, whichever agent you use:
npx skills add publora-team/publora-post-ideas
That's the skills CLI. It detects your
agent and puts the skill where that agent looks for it, so it covers Claude
Code, Codex, Cursor and Copilot without you choosing a path. Add --global to
install for every project instead of the current one.
If you'd rather not run someone else's installer, clone it and copy the folder:
git clone https://github.com/publora-team/publora-post-ideas.git /tmp/ppi
cp -r /tmp/ppi/skills/publora-post-ideas ~/.claude/skills/
For a single project, use .claude/skills/ inside the project instead.
cp -r /tmp/ppi/skills/publora-post-ideas ~/.codex/skills/
For a single project, use .codex/skills/. Restart Codex afterwards so it
reloads skill metadata.
SKILL.md is the open Agent Skills format, so the same folder works anywhere
that reads it. Copy skills/publora-post-ideas into that host's skills
directory and restart it.
I have run this in Claude Code. The skill is plain Markdown with one reference file and no scripts, so there is nothing platform-specific in it to break elsewhere, but the other hosts are untested by me.
It works on its own. The angle library is a text file and the draft comes back as text you can paste anywhere.
Connected to Publora, it also reads what you have actually published, so the angles are chosen against your real feed instead of in the abstract, the rotation is real rather than guessed, and the finished post can go into a queue instead of your clipboard. Publora's Starter plan is free forever and includes the API and MCP access this uses.
{
"mcpServers": {
"publora": {
"type": "http",
"url": "https://mcp.publora.com/mcp",
"headers": { "Authorization": "Bearer <PUBLORA_API_KEY>" }
}
}
}
The key comes from the Publora dashboard. Clients that read the official MCP
registry find the same server as com.publora/mcp-server.
Publora also publishes a set of per-network publishing skills: publora/skills covers LinkedIn, Instagram, TikTok, Threads, X, Bluesky, Telegram and a general one. This repository is the part that happens before those: deciding what to say.
MIT