ranjanvaishnav08

LinkedIn Post Skill — AI skill for Claude Code

AI community

Claude skill that drafts LinkedIn posts in your own voice from 21 hook formulas, blocks drafts that leak employer or client numbers, strips invisible AI characters, and scores each draft against your.

How to install LinkedIn Post Skill

This entry records only its repository, not the path inside it, so there is no exact command to give. Open ranjanvaishnav08/linkedin-post-skill and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What LinkedIn Post Skill does

Alternatives in AI

Invisible Dots — Open-source, self-hosted alternative to OpenAI Dots, Meta Muse, Grok Bot, Manus Cue and Claude Cowork 31.8k ★
Pixel Agents — Compressed Reference — VS Code extension with embedded React webview: pixel art office where AI agents (Claude Code terminals) are an 6.5k ★
DontFeedTheAI — Transparent anonymization proxy for AI-assisted pentesting 654 ★

README

li-post

A single Claude skill that turns one true, specific idea into a LinkedIn post in your own voice, then puts the draft through a three-part gate before you see it:

Confidentiality (scripts/guard.py): blocks a draft that names your employer, a client or a codename in the same sentence as a number, plus ticket IDs, internal URLs and "not public yet" language.
Cleanup (scripts/clean.py): auto-fixes only what cannot change meaning (invisible characters, curly quotes). Dashes and stock phrases are flagged for you to rewrite, never mangled by a regex.
Voice match (scripts/voice.py): scores the draft against your own past posts, not against a generic "sounds like AI" detector.

It never posts to LinkedIn. It writes; you post.

Install

Claude Code:

git clone https://github.com/ranjanvaishnav08/linkedin-post-skill.git ~/.claude/skills/li-post

claude.ai: download this repo as a zip and upload it as a skill.

Python 3.8+, standard library only. No network calls.

Your data stays local

Put your own files in the skill folder or `~/.claude/linkedin/`:

past_posts.txt: 5 to 10 of your posts, separated by a line with ---
guard.txt: names that must never sit next to a number, one per line (employer: ..., client: ..., codename: ...)

Both are in `.gitignore`. Do not commit them.

Limits, stated plainly

The guard is a tripwire, not legal review. An internal number with no company named in the sentence passes it.
The voice score is rough with fewer than 5 sample posts.
The hook notes in hooks.json are the original author's rules of thumb, not measured results.

Tests

python3 -I tests/test_scripts.py

Credit

Derived from [linkedin-agent-skill](https://github.com/avi691/linkedin-agent-skill) by Jake Schincariol, shared by Avi Grondin, under the MIT License. The hook formulas (`hooks.json`) and stock-phrase list (`scripts/slop.json`) come from that project. The g