Guide · Agent skill
How to use the Write Like Meng on X agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The write-like-meng-on-x skill is an agent skill that writes, rewrites and reviews X/Twitter posts, replies, threads and quote posts in Meng To's voice. It is for people who have Meng To's permission and access to his Content repo, and for anyone who wants to see how a specific author structures hooks and endings. After this guide you will know what the skill does, how to install it, how to trigger it, and what it needs before it runs.
Write Like Meng on XMengTo/Skills · MIT Download the skill Quick start
- 1Download and unzip. Download the write-like-meng-on-x ZIP from our page and unzip it. Confirm that SKILL.md and the references and scripts folders are present.
- 2Place the skill folder. Move it into ~/.claude/skills/ for personal use or into the project's .claude/skills/ folder. Alternatively, upload it under Settings → Capabilities → Skills in the Claude apps.
- 3Confirm the Content repo. Make sure you have a checkout that contains AGENTS.md, references/tweet-corpus.jsonl, voice-profile.md and content-source-map.md.
- 4Make a drafting request. Ask in plain language, for example to improve a tweet in Meng's tone, and include your draft or the subject.
- 5Review the five options. Read the labelled options, pick one, and edit it. Check every factual claim against your own knowledge before posting.
- 6Optional: refresh the corpus. Only if you need new voice evidence, sign in to X in the Codex in-app browser and ask for a corpus refresh. Node.js is required for the ingest script.
What is an agent skill and what does this one do?
An agent skill is a folder of instructions and files that an AI agent loads when a request matches the skill's description. This one is built around Meng To's writing. Once it is active, the agent drafts or rewrites posts so they sound like him, using a saved voice profile and a deduplicated archive of his authored posts.
For every drafting request, the skill returns five distinct options, with the strongest first. It also checks whether a new angle repeats something he has already said, and it grounds claims in his real product and resource context from the Content repo.
- Writes original posts, replies, threads, resource shares and quote posts as Meng
- Rewrites a draft you already have, keeping your strongest line when it sounds natural
- Checks whether an angle repeats an earlier post
- Can refresh the voice corpus from his signed-in X profile when you ask it to
Install the skill from the ZIP or the Claude apps
Download the ZIP for write-like-meng-on-x from our page and unzip it. The skill folder should contain SKILL.md, agents/openai.yaml, references/, and scripts/. Then place it in one of two locations:
For a personal install, unzip it into ~/.claude/skills/. For a single project, unzip it into the project's .claude/skills/ folder. You can also upload the skill in the Claude apps under Settings → Capabilities → Skills.
- Personal: ~/.claude/skills/
- Per project: .claude/skills/ inside the project
- Claude apps: Settings → Capabilities → Skills, then upload
Set up the Content repo the skill depends on
The skill works from a Content repo checkout. The skill expects that checkout to contain AGENTS.md, references/tweet-corpus.jsonl, voice-profile.md and content-source-map.md. If you do not have access to that repo and its bundled corpus, the skill cannot ground its drafts or check voice evidence, so it is not useful on its own.
The skill reads AGENTS.md and checks git status before it changes any files. If no Content repo is active, it will try to locate one in the workspace or ask you for the target.
- Required: Content repo checkout with AGENTS.md, references/tweet-corpus.jsonl, voice-profile.md and content-source-map.md
- Optional, only for corpus refresh: Node.js, and the Codex in-app browser signed in to X
How to trigger the skill with natural requests
You do not need a special command. Describe the task in plain language and mention Meng or his voice. The skill's description covers rewriting a tweet in his tone, drafting an original post, a reply, a thread, a resource share or a quote post, checking for repeated angles, and studying his style.
Give the skill the raw material. A draft, a link or a product detail helps it ground the post. If you want a different number of options, say so, since the default is five.
- "Improve this tweet in Meng's tone: [your draft]"
- "Draft an original post as Meng about [a product he uses]"
- "Write a quote post as Meng on this resource: [link or name]"
- "Check whether this angle repeats something Meng already posted"
- "Study Meng's writing style and update the voice profile"
Examples of what you get back
For a rewrite request, expect five labelled drafts. The first is usually the closest polish that keeps your original thought. The others vary in hook, tone, proof and playfulness. The skill is designed to avoid five near-copies of one sentence.
Each draft is written as Meng, so check it before you post it. The skill never posts anything itself.
- Option 1: closest polish, keeps your original phrasing
- Option 2: punchier hook or tighter structure
- Option 3: more conversational and spoken
- Option 4: proof-led or practical
- Option 5: playful, contrarian or use-case-driven
Safety notes: scripts and network access
This skill runs scripts and uses the network. The corpus refresh step runs scripts/update-tweet-corpus.mjs with Node.js, and it collects posts from X through the Codex in-app browser. The skill is set up to keep X read-only: it should not post, reply, like, repost, follow, DM or change account state.
The skill also stops rather than falling back. If X is logged out, blocked by a CAPTCHA, or the browser cannot attach, it reports the blocker and does not use public search, Chrome or guessed posts. Review the script and the diff before you commit any changes it makes.
- Runs scripts: yes (scripts/update-tweet-corpus.mjs)
- Uses network: yes, for the optional corpus refresh
- The refresh step needs a signed-in Codex in-app browser
- The skill never posts to X itself
node <skill-dir>/scripts/update-tweet-corpus.mjs --input <batch-path>Tips for better drafts
Give the skill a specific subject, such as a product, workflow or resource, rather than a general topic. Grounded posts sound more like Meng and are less likely to invent claims. Keep the original line you like in the request, because the skill is built to preserve it when it is already natural.
If the skill returns a draft that repeats a familiar line, ask it to check for repeated angles against the corpus. Expect it to cut any sentence that does not add a new reason, mechanism, proof point or conclusion.
- Name the product, resource or result you want to mention
- Paste your strongest line and ask for polish, not replacement
- Ask for a repeat check before you publish
- Ask for a different format if the first output is in the wrong shape
What people use it for
- Rewriting a rough tweet about a tool Meng uses so it keeps his phrasing but gets a sharper hook
- Drafting a thread that walks through a workflow from the Content repo, with a proof point
- Writing a reply in a conversational register that matches how Meng replies
- Checking whether a new resource share repeats a recommendation he already made
- Studying how an author structures hooks and endings by reading the voice profile
Frequently asked questions
How do I install the write-like-meng-on-x skill?
Download the ZIP from our page and unzip it. For personal use, put the folder in ~/.claude/skills/, or for one project, in the project's .claude/skills/ folder. You can also upload it in the Claude apps under Settings → Capabilities → Skills.
Is the write-like-meng-on-x skill free to use?
The skill is published under the MIT licence, as stated on the item page. The skill still depends on a Content repo checkout and its bundled corpus, which you need to have access to.
What do I need to run this skill?
You need a Content repo checkout containing AGENTS.md, references/tweet-corpus.jsonl, voice-profile.md and content-source-map.md. Node.js and a signed-in Codex in-app browser are needed only for the optional corpus refresh.
Does the skill post to X for me?
No. The skill is set up to keep X read-only and never posts, replies, likes or reposts. You copy and publish the draft yourself after reviewing it.
Why does the skill stop instead of using public search when X is logged out?
The skill is designed to stop and report the blocker rather than fall back to public search, Chrome or guessed posts. This keeps the voice evidence limited to real authored posts.
Can I use this skill to write posts that impersonate Meng To without his review?
No. The skill is meant for drafts written with Meng To's permission and context, and each draft should be checked before posting. It is not intended for publishing or impersonating him without his review.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you have Meng To's permission and access to his Content repo, and you want drafts that match his voice and are grounded in his real product and resource history. Pick something else if you do not have the repo and bundled corpus, or if you want a general-purpose copywriting tool for your own voice.
Download the skill