PARTLY

Partly true: the free 11-skill LinkedIn repo is real and does write posts, comments and replies, but it never touches your LinkedIn account, its 'five AI detectors' are five local scoring heuristics, and 'Claude's watermark' is an invisible-character cleanup

Facebook · Oct 2026

The claim'Claude just KILLED LinkedIn! You can now connect Claude to LinkedIn and it can basically manage your entire account for you. It's called a LinkedIn agent skill and it's completely free.' 11 skills: posts from 21 hook formulas, comments on other people's posts for you, replies, profile optimizer, weekly planner, and a humanizer that 'removes em dashes, gets rid of those obvious AI style words, removes Claude's watermark, and then runs your content through five AI detectors so it doesn't get flagged.' 'Follow me and comment X to get the setup guide.' USA/UK/Canada only.

The repo on screen is real: github.com/jakeschincariol/linkedin-agent-skill, MIT licensed, about 1.1k stars, described as 'eleven free Claude skills that run a LinkedIn account' with 'no signup, no API key, nothing to connect'. The eleven skills (li-post, li-comment, li-reply, li-profile, li-plan, li-human, li-carousel, li-repurpose, li-dm, li-inbox, li-audit) match the reel's list, and the 21 hook formulas, 12-part profile rubric and weekly plan are all in the README. From there the reel inflates everything. 'Connect Claude to LinkedIn' and 'manage your entire account': the README says the opposite, 'Nothing gets posted until you say yes. These skills write. You post.' There is no connection; you paste in a post or a comment thread and Claude drafts text. The README even warns that 'automating the site with a browser or a third-party tool violates LinkedIn's User Agreement and gets accounts restricted', which matches LinkedIn's own agreement banning bots that 'create, comment on, like, share' posts. 'Comments on other people's posts for you' is drafting, not posting. The 'five AI detectors' are, in the repo's own words, five local heuristics (burstiness, specificity, slop density, fingerprint, voice) that 'are not GPTZero, Originality, Copyleaks, Winston or Turnitin, they do not call those services, and they cannot promise those verdicts.' 'Removes Claude's watermark' refers to stripping invisible characters, em dashes and curly quotes; I found no Anthropic documentation of a text watermark in Claude output. The sell is follow-plus-comment for a 'setup guide' to a repo whose README already is the setup guide.

What holds up

Reel read frame by frame: GitHub page for 'linkedin-agent-skill' with the README line 'Eleven free Claude skills that run a LinkedIn account... Free, MIT, no signup, no API key, nothing to connect'; mock UI cards for Post Writer, Comment Drafter, Reply Handler, Profile Optimizer, Content Planner, Humanizer, Hook Extractor; a 'Checking content' card ticking Detector 1 through 5 to 'All checks passed'.
github.com/jakeschincariol/linkedin-agent-skill (fetched Oct 2, 2026): MIT, about 1.1k stars, 166 forks, 11 skills named li-post, li-comment, li-reply, li-profile, li-plan, li-human, li-carousel, li-repurpose, li-dm, li-inbox, li-audit. README: 'Nothing gets posted until you say yes. These skills write. You post.' and 'automating the site with a browser or a third-party tool violates LinkedIn's User Agreement and gets accounts restricted.'
README on the humanizer: two local Python scripts (humanize.py, detect.py) that strip invisible characters, normalize em dashes, curly quotes and ellipses, swap 113 stock phrases, then score burstiness, specificity, slop density, fingerprint and voice; 'The five checks are local heuristics, not detector APIs... they are not GPTZero, Originality, Copyleaks, Winston or Turnitin, they do not call those services, and they cannot promise those verdicts.'
LinkedIn User Agreement, section 8.2: members may not 'use bots or other unauthorized automated methods to access the Services... create, comment on, like, share, or re-share posts'. No Anthropic page documenting a hidden text watermark in Claude output was found; the repo's 'fingerprint' check is invisible characters and typography, not a cryptographic mark.

What doesn’t

'Connect Claude to LinkedIn' and 'manage your entire account' are false; the skills draft text and the README says you post it yourself.
'Runs your content through five AI detectors' is false; the repo's five checks are local heuristics that explicitly do not call any detector service.
'Removes Claude's watermark' describes stripping invisible characters and em dashes; no documented Claude text watermark exists to remove.
Follow-and-comment hook for a 'setup guide' that is already the public README; the USA/UK/Canada line is audience targeting, not a product limit.

The catch

The repo is a solid free prompt pack for drafting LinkedIn content inside Claude and worth installing, but it is a writer, not an agent, and the one thing the reel promises (Claude running your account) is the thing the author warns will get your account restricted. The detector card at the end is a scoreboard of the repo's own heuristics; it tells you nothing about what LinkedIn or a real detector will think.

How to actually do it

Install it straight from the README (clone into ~/.claude/skills or use the plugin marketplace commands); skip the comment funnel.
Use li-post and li-reply as drafting tools: feed them your own stories and numbers, edit the output in your voice, and post by hand. Do not wire any browser automation to LinkedIn.
Treat li-human as a style linter, not a detector pass. If a client or employer uses a real detector, the only reliable defense is writing that is actually yours.

The free MIT repo with eleven Claude skills exists and matches the reel's feature list, so the tool is real. The reel's headline claims fall apart on the README itself: nothing connects to LinkedIn, nothing posts, the five 'detectors' are local heuristics, and the author warns that automating LinkedIn gets accounts restricted.

Confidence: High
Posted by: an AI-tips creator posting Claude tutorials (about 1k views, 2.2k reactions on this reel); reel sent to Buddy Tue Sep 29, 2026, ~2:07 PM

See the original claim →

We test hype for free. We build the real thing for a living.

Thirty minutes, no pitch — and you'll leave with something useful either way.

Book a call with Todd or start with the free Business Checkup →

The Verdict Weekly

Three verdicts every Friday. Free forever, unsubscribe anytime, no spam — that would be ironic.