LinkedIn Agent Skill
A pack of eleven MIT-licensed Claude skills to prepare what you publish on LinkedIn: `/li-post` (a post from 21 hook formulas), `/li-comment`, `/li-reply`, `/li-profile` (your profile scored on a 12-part rubric), `/li-plan` (the week on one page: what to post, on which day, and the ten people to engage with, written to `~/.claude/linkedin/plan.md`; that is the 'who to engage with' promised by the [Instagram reel DdNAFIZNIt7](https://www.instagram.com/reel/DdNAFIZNIt7/), a list, not comments sent on your behalf), `/li-human`, `/li-carousel`, `/li-repurpose` (one piece turned into a week of posts), `/li-dm`, `/li-inbox` and `/li-audit` (your published posts ranked by engagement). `/li-human` strips invisible characters, long dashes, curly quotes and 113 stock phrases, then scores the text on five local checks: sentence variation, specificity, stock-phrase density, typographic fingerprint and voice. Nothing connects to LinkedIn: you paste the text yourself.
Strengths
- `/li-human` flags in one pass the marks of a generated text (invisible characters, stock phrases, long dashes)
- No access to your account: nothing is posted unless you paste the text yourself
- Free, MIT, installed by pasting the repo link into Claude
Limitations
- Hook formulas standardise the voice: your post looks like the ones it imitates
- The humanity score measures tics, not the value of what is said
- A young project (a few hundred stars) maintained by one person
Best for
- Reviewing a hand-written post before scheduling it, to hunt down generated-text tics
- Auditing your LinkedIn profile against an explicit rubric
- Ranking your already published posts by engagement to see what actually landed