Write, edit, and approve LinkedIn posts from the terminal using 12 skills
You can draft posts for your LinkedIn feed, leave comments, and refine your profile using terminal commands. This repository contains 12 skills that guide Claude Code and Codex through LinkedIn marketing tasks, helping to remove the awkwardness often found in AI-generated text and produce a more human voice. Every action requires you to review a draft and give explicit approval before it is executed, preventing accidental publication of incorrect content.
Removing AI traces and pre-publication approval
These skills go beyond simple writing by applying rules that exclude words frequently used by AI (such as leverage and fundamentally) and emphasize specific numbers and experiences. The Humanizer skill specifically assigns an AI vocabulary score based on paragraph density to ensure readers do not perceive the text as machine-generated. All skills generate a draft and wait for user confirmation; nothing is posted to LinkedIn without approval. This design maintains the efficiency of automation while keeping final responsibility with the user.
The role of the 12 skills
The repository includes 12 skills that handle various tasks, from writing to data analysis and profile optimization.
- Writing and editing Post Writer uses 20 verified hook formulas to create posts with viral potential, while Humanizer removes AI-specific awkwardness. Repurposer converts content from other platforms into a format suitable for LinkedIn.
- Interaction management Comment Drafter and Reply Handler write comments and replies for specific posts, accurately handling LinkedIn's two-stage thread structure. Engagement Monitor tracks comment threads and categorizes followers according to target customer groups.
- Strategy and analysis Content Planner creates a 7-day content plan, and Hook Extractor reverse-engineers the hook formulas of viral posts. Profile Optimizer rewrites the headline and self-introduction sections to align with 2026 conversion patterns.
- Data integration Integrating Apify allows you to automatically retrieve LinkedIn data, including post bodies, comments, and people who liked posts. Connecting Publora enables you to publish drafts directly to LinkedIn or schedule them.
Installation and requirements
The skills can be used in various agent environments, including Claude Code, Codex CLI, Claude Desktop, and OpenClaw. If you use Claude Code or Codex, add sergebulaev/linkedin-skills from the platform's plugin marketplace or clone the repository to install it. In the Claude.ai web version or desktop app, you can add it from the marketplace via the Plugins tab in the 'Customize' menu. To use data reading (Apify) and automatic posting (Publora), you must set their respective API tokens in the .env file; image generation (Pixfaro) is optional. In Windows environments, be careful with symbolic link settings, as enabling developer mode or setting git config core.symlinks true is required.
Pre-use considerations
This repository is released under the MIT license, allowing free use. However, the automatic data reading feature via Apify is not supported in the Claude.ai web version, in which case you must paste the relevant text manually. Publora's free plan supports 15 posts per month, and Apify's free credits are worth $5 per month, which is sufficient for typical usage. This tool provides the 'content engineering' layer of LinkedIn marketing, focusing on building data collection, engagement management, and posting infrastructure rather than selling the writing itself. It is suitable for founders or marketing professionals who want to systematize their LinkedIn activity while maintaining their own voice.