x-agent
X Agent
The front door: routes any X request to the right skill, sets up the voice file first, and suggests a weekly routine built around replies.
npx skills add liftli-ai/x-agent-skills --skill x-agentX (Twitter) agent skills
11 free, MIT-licensed X agent skills that turn Claude Code, Codex, Cursor or any skills-compatible agent into an X content team. Call it Claude for X, or Claude for Twitter: posts, threads, hooks and replies in your voice, LinkedIn posts rebuilt for X, a better bio, and a check of every draft against what X's own ranking code rewards. Your agent's own model runs them: no X API key, no account, nothing phones home.
npx skills add liftli-ai/x-agent-skillsOr as a Claude Code plugin:
/plugin marketplace add liftli-ai/x-agent-skills/plugin install x-agent-skills@liftli-x● MIT license · runs on your agent's model · github.com/liftli-ai/x-agent-skills
What these are
A skill is an installable instruction package: a SKILL.md your agent reads and runs with its own model. Each of these teaches your agent one X job done well: a post that makes sense cold on a phone, a thread where post 1 works alone, a reply worth answering, a bio a stranger understands in one read.
New here? Ask your agent "help me with my X". The x-agent skill picks the right skill, and x-voice builds a voice file from your own posts and replies so drafts sound like you. It is the same file the LinkedIn agent skills read, so you build it once. The skills write and check; they never post. x-publish hands you a copy-ready post, or publishes through Liftli after you approve.
Install everything with the command above, or a single skill with --skill <name>. Every card below carries its exact command.
Start here The front door, your voice file, and publishing
x-voice
X Voice File
Builds or updates the shared voice file, .liftli/voice.md, from the user's own posts and replies. It is the same file the LinkedIn skills read, plus a section for how the user sounds on X.
npx skills add liftli-ai/x-agent-skills --skill x-voicex-publish
X Publish
Gets a final post live: a copy-ready post or thread to paste, or publishing and scheduling through Liftli, via X's official API, after the user approves. Never scripts the X website.
npx skills add liftli-ai/x-agent-skills --skill x-publishWrite Posts, threads, hooks, replies and repurposing
x-post-writer
X Post Writer
Single posts in two lengths: 280 characters counted X's way (emoji count 2, every link 23), or long posts up to 25,000 characters for Premium accounts. Three angles, each with its count.
npx skills add liftli-ai/x-agent-skills --skill x-post-writerx-thread-writer
X Thread Writer
Numbered threads where post 1 stands alone and every post earns the next, in story, teardown or list form. Says when a single post or a long post would do the job better.
npx skills add liftli-ai/x-agent-skills --skill x-thread-writerx-hook-generator
X Hook Generator
8 first lines labeled by pattern, each checked against three gates: clear with zero context, real stakes, and an open loop worth replying to.
npx skills add liftli-ai/x-agent-skills --skill x-hook-generatorx-reply-generator
X Reply Generator
Short, specific replies of 5 to 25 words that the user edits and posts. Explains, from X's published ranking code, why replies are the growth lever, and bans the tells that get replies muted.
npx skills add liftli-ai/x-agent-skills --skill x-reply-generatorlinkedin-to-x
LinkedIn to X
Rebuilds a LinkedIn post for X instead of pasting it: a standalone post under 280 characters, a short thread, and a long post for Premium. Strips the LinkedIn habits that read as imported.
npx skills add liftli-ai/x-agent-skills --skill linkedin-to-xx-to-linkedin
X to LinkedIn
Expands an X post or thread for LinkedIn: the context, the number and the story put back, sized for the see-more fold, without corporate filler.
npx skills add liftli-ai/x-agent-skills --skill x-to-linkedinCheck The pre-publish check, grounded in X's code
x-algorithm-check
X Algorithm Check
A 12-point pre-publish check. Each point is tagged as backed by X's open-sourced ranking code (replies, quotes, shares, follows; mutes, blocks, reports) or as craft. Lists the popular algorithm claims that have no source.
npx skills add liftli-ai/x-agent-skills --skill x-algorithm-checkProfile Bio, name, pinned post
x-bio-generator
X Bio Generator
Five bios within X's 160-character limit that tell a stranger who the user helps and why to follow, plus a name, a pinned-post idea and header text.
npx skills add liftli-ai/x-agent-skills --skill x-bio-generatorWhy the X skills start with replies
X ranks the For You feed by predicting what each reader will do with a post and adding up weighted predictions. The skills follow what X has published, not folklore:
- Conversation outweighs approval. In the ranking code X released in 2026, a predicted reply or quote is weighted 5.0, a like 0.6 (xai-org/x-algorithm). The 2023 release weighted a reply 13.5 and a reply the author engages with 75 (the-algorithm-ml).
- Bad posts cost. Mute author is weighted -58.8, not interested -47.52, block -31.2 and report -234. Bait and rage hooks risk exactly those.
- Read the weights honestly. X notes that they scale predicted probabilities, not raw counts. They show direction, never a "one reply equals N likes" sum.
- A human posts every reply. X's automation rules forbid automated replies based on keyword searches, and AI reply bots need X's prior written approval. So the skills draft and you post.
x-algorithm-check carries the full weight table, the date it was checked, and the popular claims (such as "links cut reach 50%") that the published code does not support.
Skills vs. the full pipeline
Skills are stateless: your agent brings the model, the skill brings the craft. What they can't know is you: your voice, your strategy, what you shipped this week, what worked last month. That is the Liftli strategist for X. It connects to the same agent as an X MCP server, learns your voice from your real posts, keeps your strategy between chats, and publishes or schedules posts and threads through X's official API after you approve each one. X publishing is on the Business plan, with a fair-use cap of 100 posts a month. Connector details for agents: liftli.ai/llms.txt.
X agent skills on GitHub, compared
A few public X skill collections exist on GitHub, and they do different jobs. Star counts are rounded, as of 14 October 2026.
| Repository | Stars | What it covers | Install |
|---|---|---|---|
| liftli-ai/x-agent-skills (this page) | new | 11 skills: a router, a shared voice file, posts, threads, hooks, replies, LinkedIn and X repurposing, a bio writer and a pre-publish check tied to X's published ranking code. No API key. MIT. Optional publishing through Liftli, via X's official API, after you approve. | npx skills add or Claude Code plugin |
| rohunvora/x-research-skill | 1.2K | Research on X from the terminal: search with engagement sorting, full threads, account watchlists and sourced briefings, through the X API. Needs your own X API bearer token and Bun. No license file. | Git clone into your skills folder |
| wshuyi/x-article-publisher-skill | 870 | Turns a Markdown file into an X Article by driving X's editor in your browser through Playwright. It saves a draft and never publishes on its own. MIT. | Git clone or Claude Code plugin |
| Hao0321/claude-skill-social-post | 730 | A Codex and Claude Code skill, documented in Chinese, that learns your voice, plans content and writes posts for Facebook, Instagram, Threads and X, then publishes after you confirm. MIT. | Git clone |
Picking one: for research on what people say on X, x-research-skill. For long X Articles written in Markdown, x-article-publisher-skill. For writing posts, threads and replies in your voice, with checks grounded in X's own code and a safe path to publishing, this collection. They are plain markdown files, so you can mix them.
Also on LinkedIn
The sibling set, 31 LinkedIn agent skills, covers LinkedIn posts, hooks, comments, profile audits and pre-publish checks. Install it with npx skills add liftli-ai/linkedin-agent-skills. Both sets share one voice file.
Frequently asked questions
What is an X agent skill?
An X agent skill is a SKILL.md instruction package that teaches an AI agent such as Claude Code, Codex or Cursor one X (Twitter) job: writing a post or thread, drafting replies, fixing a bio or checking a draft before it goes out. The agent's own model runs it, so there is no X API key and nothing to connect. This page lists 11 of them, free and MIT-licensed.
How do I use Claude for X (Twitter)?
Install these skills into Claude Code with npx skills add liftli-ai/x-agent-skills, or as a plugin with /plugin marketplace add liftli-ai/x-agent-skills and then /plugin install x-agent-skills@liftli-x. Then ask your agent for help with your X. The x-agent skill routes the request, and x-voice builds a voice file so drafts sound like you. To let Claude publish to X, connect Liftli, which posts through X's official API after you approve.
Can Claude Code post to X?
Not with skills alone, and it should not script the X website: X's automation rules forbid non-API automation and say it can lead to permanent suspension. Posting needs an app connected to your account through X's official API. The x-publish skill gives you a copy-ready post or thread, or connects Liftli, which publishes and schedules through the official API after you approve each post. X publishing is on Liftli's Business plan.
Why do these skills focus on replies?
Because X's own ranking code does. In the For You algorithm X published in 2026, a predicted reply or quote is weighted 5.0 against 0.6 for a like, and mutes, blocks, reports and not-interested clicks carry large negative weights. X notes that the weights scale predicted probabilities, not raw counts, so they show direction, not arithmetic. The x-reply-generator skill drafts short, specific replies that you post yourself.
Do I need an X API key or a developer account?
No. The skills are markdown files that your agent's model reads, and they never call X. You post the result yourself, or connect Liftli, which handles X's official API for you.
Do these work with the LinkedIn agent skills?
Yes. Both sets read the same voice file, .liftli/voice.md, so you build it once; the X set adds a section for how you sound on X. linkedin-to-x and x-to-linkedin move a post between the two networks. The 31 LinkedIn skills are at liftli.ai/skills.
Which AI agents support these skills?
Any client that reads the open SKILL.md format, such as Claude Code, Codex and Cursor. Install with npx skills add liftli-ai/x-agent-skills. The skills are plain markdown, so an agent without a skills manager can read them straight from the GitHub repository.
Are the skills really free?
Yes. They are MIT-licensed and run on your agent's own model, so Liftli sees nothing and meters nothing. They are the single-job versions of what Liftli does as a full pipeline: voice learned from your real posts, strategy memory, and publishing to X and LinkedIn behind your approval.