--- name: x-post-writer description: "Draft a single tweet or short auto-thread for X (Twitter) using a 2026 X hook formula (one-liner contrarian, data-point, build-in-public, mini-list, relatable cold-open), picked by goal (replies, reposts, likes, bookmarks). Respects the 280-char limit (25,000 on Premium), runs the humanizer pass, and publishes via Publora on approval. Use to write a tweet from notes. Not for long threads (use x-thread-builder) or auditing a draft (use x-humanizer --mode audit)." --- # X Post Writer Ship a single tweet, or a tight 2-3 tweet micro-thread, using hook shapes that actually travel on X in 2026. Single-tweet formulas land a complete punch in the first line, because X has no "see more" fold. ## When to use - User says "write me a tweet about X" - User has a topic or a rough one-liner and wants a sharper hook - User wants to pick a proven single-tweet shape and fill in their voice - User wants a quick draft that auto-publishes on approval ## Formulas this skill uses (single-tweet shapes) | Code | Formula | Primary goal | Best for | |---|---|---|---| | X1 | One-Liner Contrarian | reposts | a sharp, defensible against-the-grain take | | X2 | Data-Point Hook | bookmarks | one odd-precision number that reframes a thing | | X3 | Build-in-Public Confession | replies | a real metric from your own work, ugly ones included | | X4 | Quote-Tweet Dunk-with-Value | reposts | adding a layer to someone else's take (quote tweet) | | X5 | Mini-List Tweet | bookmarks | a scannable list that fits in one tweet | | X6 | Relatable Cold-Open | likes | a specific shared moment, no setup | | X11 | Third-Person Case-Study Flex | reposts | someone else's absurd, specific result as anyone-could-copy bait | Full skeletons in `../../references/hook-formulas.md`. For long-form tweetstorms (X7-X10), use `x-thread-builder`. X11 was validated as the top-recurring winning shape in a ~450-tweet corpus (July 2026); prefer it over a first-person "How I" when the result you are citing is someone else's. ### Pick by goal first | Goal | Reach for | |---|---| | Replies | X3, X1 | | Reposts | X1, X4, X11 | | Likes | X6 | | Bookmarks | X2, X5, X11 | ## 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, angle, any rough draft, target audience (builders / founders / a niche), and the goal (replies / reposts / likes / bookmarks). 2. **Pick the container.** If the idea is one claim, one number, or one moment, it is a single tweet. If it is a scannable one-line-per-item list, still a single tweet (X5). If items need teaching or the idea has a build, hand off to `x-thread-builder`. 3. **Pick the formula.** Use the goal table to shortlist, then suggest 2-3 that also fit the topic and let the user choose. 4. **Draft the tweet.** Fill the skeleton in the user's voice. Respect the 2026 X rules: - First line carries the whole load (no fold). - Under 280 chars on a standard account (emoji count as 2 chars). Confirm the user's tier; only go long if they are on Premium and the length earns it. - One idea per tweet. Line breaks as beats. - 0-1 hashtag at the end, 0-1 emoji, none on a serious take. - No external link in the tweet (offer to put it in a reply). 5. **Humanizer pass.** Scrub 2026 AI vocab by density, cap em dashes (at most one per tweet, never swap one for a period), break stacked triads, generic openers and reveal bridges. Leave a single tweet's rhythm alone (uniform rhythm wins on short posts; never insert a fragment for punch). Add an odd-precision number with a named referent or a named entity where the claim allows it. Canonical rules: `x-humanizer` V3. 6. **Optional audit.** Invoke `x-humanizer --mode audit` for a pass-fail check. 7. **Approval card.** Show: formula used, full draft, char count (note if emoji push it over 280), suggested posting window, primary goal. 8. **On approval.** Call `lib.publish(kind="post", draft_text=, target_url="https://x.com/compose/tweet", platforms=[], scheduled_time=)`. The wrapper handles Publora / manual / diy routing. If the content runs long, Publora auto-splits it into a numbered thread. ## Hard rules Global voice rules: see root `SKILL.md` Voice rules. Additional skill-specific rules: - The first line must stop the scroll on its own. Rewrite any opener that needs line 2 to make sense. - Keep a standard-account tweet inside 280 chars. If it does not fit, tighten it or escalate to a thread, never ship a truncated thought. - One specific number where the claim allows it. 2.4x beats "way better". - Do not hard-sell the user's product. One natural mention max. ## Anti-patterns (skill will refuse) - ALL CAPS first line for intensity. Carry intensity with word choice. - Em dashes above the cap (more than one in a tweet), or an em dash swapped for a period. - "Here's what / here's how" or "Stop X, start Y" as the opener; "The result?" as a reveal. - Announced candor ("Let me be honest", "not gonna lie") with no dated fact behind it. - "Unpopular opinion:" on a take that is actually popular. - Padding a one-line idea with filler to look substantial. - Rule-of-three lists without specifics. - "leverage", "fundamentally", "game-changer", "deep dive". - External link in the tweet body. - Engagement bait ("RT if you agree", "reply YES"). ## Resources - `../../references/hook-formulas.md` - all 11 X formula skeletons (X1-X6 and X11 are single-tweet) - `../../references/algorithm-heuristics.md` - 2026 X ranking rules (signals, timing, limits) - `references/single-tweet-checklist.md` - the per-tweet scrub and fit list ## Optional illustration Offer a generated image when a visual would lift reach. Draft a prompt and call `lib.illustrate(prompt, kind="wide")`, pulling brand handle/color from Voice & Brand Profile section 6 for a pixel-exact overlay. Show the returned `url` + `cost`, then attach on publish via `media_urls=[url]`. Full workflow (incl. quote-cards): `../x-humanizer/sub-skills/illustration.md`. No Pixfaro key -> it drafts the prompt for you to generate manually. ## Related skills - `x-thread-builder` - when the idea needs more than a tweet - `x-humanizer` - aggressive AI-tell scrubber, plus `--mode audit` for review - `x-hook-extractor` - reverse-engineer a hook from a tweet you admire