write-like-meng-on-x

Drafts and refines X posts in Meng To's voice using a deduplicated tweet corpus.

5.7k|685|Updated Feb 3, 2026
One-click install
npx skills add https://github.com/MengTo/Skills --skill write-like-meng-on-x
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: write-like-meng-on-x
Source: https://github.com/MengTo/Skills/tree/main/agent-skills/codex/write-like-meng-on-x
Command: npx skills add https://github.com/MengTo/Skills --skill write-like-meng-on-x

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

Writing authentic X/Twitter posts in a specific person's voice is hard without evidence of how they actually write, and it is easy to repeat old hooks or invent facts. This Skill grounds every draft in Meng To's deduplicated authored-post corpus, a curated voice profile, and verified personal and product context.

Core Features & Use Cases

  • Voice-grounded drafting and rewriting: Produces exactly five materially different options for any tweet, reply, quote post, thread, or resource share, matched to the format's natural style.
  • Duplication prevention: Searches the tweet corpus by ID, URL, and text hash to avoid repeating earlier hooks, claims, anecdotes, or endings.
  • Corpus refresh workflow: Collects 20-50 new authored posts via the Codex browser and ingests them with a validation script that rejects duplicates and malformed records.
  • Use Case: Ask the agent to rewrite a product announcement tweet in Meng's tone; it reads the voice profile, checks the corpus for similar past posts, and returns five grounded options without inventing metrics or facts.

Quick Start

Use the write-like-meng-on-x skill to rewrite this tweet in Meng's current voice and give me five options.

Frequently Asked Questions about write-like-meng-on-x

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I write tweets in someone else's voice with AI?

Build a corpus of the person's authored posts and a voice profile documenting their tone, formats, and recurring themes. This Skill reads those references before drafting and returns five distinct options grounded in verified facts rather than generic copywriting templates.

How does the skill avoid repeating old tweets?

It checks both corpus duplication and editorial duplication. The update-tweet-corpus.mjs script matches posts by ID, canonical URL, and normalized text hash, while the drafting workflow searches for repeated hooks, claims, anecdotes, and endings before finalizing.

Can I use this skill with Claude or Cursor instead of Codex?

The drafting and review instructions are plain Markdown and work with any agent that can read SKILL.md and the references folder. However, the corpus refresh workflow specifically requires the Codex in-app browser for read-only X access.

What happens if X is logged out during corpus refresh?

The refresh stops and reports the exact blocker, such as a logged-out session or CAPTCHA. The skill explicitly forbids falling back to public search, Chrome, or guessed posts, keeping the corpus limited to verified authored content.

Does the skill invent personal details or metrics for posts?

No. The instructions strictly prohibit inventing usage, metrics, customers, travel, family details, or product capabilities. Drafts must be grounded in the tweet corpus or trusted Content repo sources listed in the source map.