threads-skills: A Claude Code Skill for Writing Threads Posts in Your Own Voice

Threads caps every post at 500 characters, drops a second hashtag without warning, and doesn't add the "(1/N)" markers X uses on multi-post threads. Those are the kinds of platform-specific rules a general-purpose writing prompt tends to miss, and they're exactly what threads-skills is built around: eight Claude Code and Codex skills that write, audit, and plan Threads content with those constraints baked in rather than bolted on.

The repository comes from developer Serge Bulaev and sits inside a small family of platform-specific skill packs, an Instagram version, an X version, a YouTube version, a TikTok version, a Facebook version, and the original LinkedIn version that's crossed 2,500 GitHub stars. threads-skills itself is newer and smaller, 40 stars as of this week, MIT licensed, free, with a commit landing the same day this piece was written. That size difference matters for how to read the traction: it's early, but it's built on the same voice engine and approval flow that the larger sibling tools have already proven out.

What threads-skills actually installs

The bundle covers eight distinct jobs rather than one general-purpose "write me a post" prompt. Post Writer drafts a single post or a full multi-post thread using one of 13 documented Threads hook formulas, chosen by the goal you state, replies, reposts, likes, or quotes, and it respects the 500-character limit along with the one-hashtag cap the platform actually enforces. Hook Extractor works the other direction: paste in a viral Threads post and it maps the hook to one of those same 13 formulas and hands back a blank template. Reply Drafter writes a reply or a value-add quote post for a given post URL. Repurposer takes a LinkedIn post, an X thread, a blog post, or a YouTube script and rewrites it as a native Threads post, stripping the tone and formatting that gives away its origin platform. Content Planner builds a week of posts with daily hooks and posting times. Profile Optimizer audits a Threads bio and pinned post against a 150-character limit and a searchable-handle check. Audience Insights pulls real engagement data for a query or a competitor's profile through Apify, without needing a login.

The eighth skill, Humanizer, is the one doing the most structural work. It scores a draft for AI vocabulary density, caps em dashes at one per post instead of banning them outright, and specifically flags uniform sentence rhythm, since the README's own claim is that consistent rhythm, not any single word choice, is what reads as machine-written on Threads at every post length. It also runs as a pre-publish audit mode, checking character count, hook strength, hashtag count, and link placement in one pass before anything goes out.

How posting and approval actually work

Every skill in the bundle drafts content and stops there by default, you copy the draft into Threads yourself. Auto-posting is optional and runs through Publora, a publishing API that solves a specific technical problem: Threads' native API builds a multi-post thread by creating each post as a reply to the one before it, which means handling partial failures yourself if step three of five fails. Publora takes a thread written as one block, splits it at paragraph and sentence boundaries, and posts the whole chain in a single call. Setup takes signing up for a free Publora account, connecting a Threads channel, and dropping an API key and platform ID into a local .env file, roughly a two-minute job according to the documentation.

One limitation is worth stating plainly because the README does: a reply to someone else's post is technically a separate post that Publora's publish call can't target, so Reply Drafter always returns a copy-paste block rather than auto-posting, even with Publora connected. The README also flags that Meta's own Threads-app reconnection process can temporarily pause multi-post nesting, in which case the fallback is posting the opening line and adding the rest as manual replies. Neither is a dealbreaker, but both are the kind of platform friction a tool like this can route around, not eliminate.

Who this fits, and who should skip it

This is a fit for someone already running an X or LinkedIn presence with Claude Code or Codex in their workflow, who wants to add Threads as a real channel without learning a third platform's unwritten rules from scratch, or for a marketer testing Threads specifically because, as of this week, it's the one major Meta surface that hasn't been folded into a paid tier. It's not a fit for anyone who wants a fully autonomous posting pipeline with zero review, the approval step is a deliberate design choice, not a missing feature, and it's not the right tool if you need to track who's replying to or DMing you, since Threads walls off the engager list entirely, Audience Insights is discovery and profile stats, not a follower roster.

The honest caveat on traction: 40 stars is a small base, and this is functionally a one-developer project extending a proven format into a new platform, not yet a widely adopted tool in its own right. The mechanics and the voice rules are specific enough to be worth trying regardless, but treat early adoption here as a reason to test it yourself rather than a signal that it's already been validated at scale the way its LinkedIn sibling has.

Source: sergebulaev/threads-skills on GitHub.