Linkedin Post Analyser

linkedin-post-analyser

A Claude Code / Claude Cowork skill that scores a drafted LinkedIn post across 7 platform-specific dimensions before you publish. Returns an /100 score, an auto-fail check, line-level fixes and a ship verdict. The final gate in a content pipeline, not a rewriter.

Why this exists

Voice tools tell you whether a post sounds like you. They do not tell you whether it works on LinkedIn. Those are different questions, and the second one is where drafts quietly die.

This skill answers the second one with a rubric instead of a vibe check:

• 7 dimensions, 10 points each, scaled to /100
• Auto-fail conditions are binary. One of them fires and the post does not ship, whatever the total says
• Line-level fixes, not paragraph feedback. Every flag names the line and the change
• A machine-readable report block written back to the draft, so any dashboard or script can trend scores over time

That last part is the point. One score is an opinion. A hundred logged scores tell you which patterns actually earn reach on your account.

First run

The skill ships with **zero identity in it**. It does not know your brand, your audience or your CTAs, and it will not guess.

On first use it runs a setup wizard (`SETUP.md`) that asks about your brands, who reads them, their pains in their own words, your voice cues, and the CTAs you actually use. It writes the answers to `config/brand-profile.md`.

`config/` is gitignored. Your answers never get committed, and never leave your machine.

Re-run the wizard any time to add a brand or update the audience. It updates the config in place and never touches the skill.

How it pairs with the family

Skill	What it does
voice-analyser	Extract: builds tone-of-voice.md from 3 real writing samples
voice-humaniser	Rewrite: strips AI-writing tells and rewrites in the doc