X Thread Builder

X Thread Builder

Build tweetstorms that get read all the way down. On X, tweet 1 is the entire funnel: it has to promise a payoff and open a loop, or no one taps "Show this thread." Everything after only matters if tweet 1 earns the tap.

When to use

• User says "turn this into a thread" or "write me a thread about X"
• The idea is a list where each item needs teaching, a story with a build, a surprising result with a mechanism, or a repeatable process
• User has a single tweet that is trying to do too much and should be a thread

Formulas this skill uses (thread shapes)

Full skeletons in ../../references/hook-formulas.md. For single tweets (X1-X6), use x-post-writer.

Pick by goal first

How Publora splits a thread

Pass the full thread as one content string. Publora auto-splits it:

• at paragraph breaks (\n\n) first,
• then at sentence endings,
• then at word boundaries as a fallback,
• and adds (1/N) markers, reserving ~8 chars per tweet for the marker.

For full control of where each tweet breaks, hand-split with a --- separator on its own line between tweets. The skill defaults to hand-splitting with --- so the user sees exactly how the thread will land before approving.

Steps

Voice profile first (all drafts). If ../../references/voice-profile.md has filled: yes, load it and match the user's voice fingerprint, hard rules, and CTA/link style throughout. If it is not filled, mention once that x-humanizer --mode profile can learn their voice from a few posts, then proceed with the generic voice rules.

Story bank next. If ../../references/story-bank.md has filled: yes, load it and take every concrete detail (numbers, dates, named projects, positions) from there instead of asking mid-draft. Never invent a figure, a date or a client name that is not in the bank or supplied by the user in this conversation. If the bank has nothing that fits, ask the user, or say once that x-humanizer --mode interview fills it in 20 to 40 minutes. A draft with a plausible invented number is worse than a draft with none: one is vague, the other is a retraction.

1. Gather inputs. Topic, the raw material (notes, a transcript, a result), target audience, and the goal (bookmarks / likes / reposts / replies).
2. Pick the formula from the goal table, confirm it fits the material.
3. Write tweet 1. This is 80% of the work:
   • Promise a specific payoff (a number, a named result, a clear list count).
   • Open a loop the reader has to resolve ("most people get 3 of them wrong", "I did not expect why").
   • Keep it under 280 chars on a standard account. It must stop the scroll alone.
4. Pace the body. One beat or one item per tweet. Front-load the strongest item or beat at position 1-2; tap-through decays with depth. Target 5-9 tweets for teaching/list threads; longer only for a strong story.
5. Make each tweet stand alone. A reader landing mid-thread from a quote tweet should still get something.
6. Write the closer. The last tweet earns the repost and the follow. End on the most quotable line, then one clear ask (bookmark, follow, or "reply with yours"), not both.
7. Humanizer pass on every tweet. Scrub 2026 AI vocab by density, cap em dashes (at most one per tweet, never swap one for a period), break stacked triads and reveal bridges. Let tweet length vary only as the material varies: never insert a 3-word punch tweet for rhythm, never alternate long/short, and touch rhythm only if the whole thread reads machine-flat. Canonical rules: x-humanizer V3.
8. Optional audit. Run x-humanizer --mode audit on the assembled thread.
9. Approval card. Show the thread tweet-by-tweet with per-tweet char counts, total tweet count, the formula, and the goal. Confirm any --- breaks.
10. On approval. Call lib.publish(kind="thread", draft_text=<full thread with --- breaks>, target_url="https://x.com/compose/tweet", platforms=[<X_PLATFORM_ID>], scheduled_time=<iso_or_None>). Publora posts the chain in one call and returns a postGroupId.