agentleFS
megandmartinmegandmartin/agent-skills-repo/skills/content-marketing/x-thread-builder/SKILL.md

x-thread-builder

Skill4 starsChanged 4 months agoMITView on GitHub

Builds an X (Twitter) thread — one scroll-stopping opener, 5–9 self-contained body tweets, a CTA close — with a python3 character-count check on every tweet. Use when the user asks for an "X thread", "Twitter thread", "thread about [topic]", "tweetstorm", or wants long-form content broken into tweets. Don't use for single tweets (draft directly), LinkedIn posts (linkedin-content-pack), or fanning one piece across many platforms (content-repurposer).

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

Files
SKILL.md

X Thread Builder

Builds threads where every tweet earns its place: an opener that stops the scroll on substance (not bait), 5–9 body tweets that each stand alone when quoted out of context, and a close with one clear CTA. Every tweet is machine-verified under 280 characters before delivery. Output is always a draft — nothing gets posted.

When to Use

  • The user wants a thread from a topic, article, brief, or set of notes.
  • A deep-research-brief or long-form piece needs an X-native breakdown.
  • Not for: single tweets (draft directly), LinkedIn posts (linkedin-content-pack), or fanning one piece across many platforms (content-repurposer).

Quick Reference

ActionCommand / Call
Count one tweetpython3 -c "print(len('''TWEET TEXT'''))"
Count whole threadWrite tweets to /tmp/thread.txt separated by ---, then run the step 5 script
URL length ruleAny URL counts as 23 chars regardless of actual length
Thread size1 opener + 5–9 body + 1 close = 7–11 tweets total
Numbering2/ style on body tweets; opener unnumbered or 1/

Procedure

  1. Precheck — confirm python3 is available (python3 --version). Confirm you have the source material, the audience, and the one thing the reader should do at the end (follow, read, reply, sign up). One CTA only — ask if unstated.
  2. Outline — list 5–9 points, one per body tweet, in an order that builds (setup → payoff, or list of escalating value). Each point must survive alone: assume any tweet may be someone's first contact with the thread.
  3. Draft the opener — a concrete claim, result, number, or tension in ≤280 chars. It must promise exactly what the thread delivers. Banned: "🧵👇" as the whole hook, "You won't believe", "Most people don't know this", curiosity gaps with no payoff, "A thread on threads".
  4. Draft body tweets — one idea each, ≤280 chars, self-contained (restate the subject; no bare "It also..." referencing a previous tweet). Line breaks inside a tweet are fine and free. Numbers, specifics, and short sentences over adjectives. No mid-thread hashtags.
  5. Character check (hard gate) — save the thread to /tmp/thread.txt with tweets separated by lines containing only ---, then run:
    python3 -c "
    import re
    ts=[t.strip() for t in open('/tmp/thread.txt').read().split('\n---\n')]
    for i,t in enumerate(ts,1):
        n=len(re.sub(r'https?://\S+','x'*23,t))
        print(f'{i}: {n} chars', 'OK' if n<=280 else 'OVER')
    "
    Expected: every line prints OK. Any OVER tweet gets trimmed (cut adjectives and hedges first) and the script rerun until clean.
  6. Close with CTA — final tweet: one-line recap of the payoff + the single CTA + at most one link. Optionally "If this was useful, RT the first tweet" — once, plainly, no begging.
  7. Deliver — output the template below with the verified character counts shown per tweet. Recommend a brand-voice-guardian pass if a voice profile exists. Do not post.

Output Template

# X Thread: [topic]
**Audience:** [who] | **CTA:** [the one action] | **Tweets:** [n]

1/ [opener]  (xxx chars ✓)

2/ [body tweet]  (xxx chars ✓)

...

n/ [close + CTA]  (xxx chars ✓)

## Notes
- [attachment suggestion for tweet 1, if any: image/screenshot/chart]
- [best posting window + timezone assumption]

Pitfalls

  • Eyeballed character counts — a "looks fine" tweet at 293 chars breaks on posting. Recovery: the step 5 script is mandatory, not optional; rerun after every edit, including "tiny" ones.
  • Dependent tweets — tweet 4 reads "It also does this" and means nothing when quoted alone. Recovery: reread each body tweet in isolation; restate the subject noun in any tweet that fails the standalone test.
  • Opener overpromises — hook says "10x your output", thread delivers three mild tips. Recovery: rewrite the opener DOWN to what the thread proves, or upgrade the weakest body tweets with concrete numbers/examples until the promise is honest.
  • Multi-CTA close — follow me, AND read the blog, AND join the list. Recovery: keep the CTA chosen in step 1; move the rest to a reply-to-self suggestion in Notes.

Verification

  • Step 5 script run on final text; every tweet printed OK (counts shown in output)
  • 7–11 tweets total; each body tweet passes the standalone read test
  • Opener contains a specific claim/number and none of the banned patterns
  • Exactly one CTA in the close; no mid-thread hashtags
  • Delivered as a draft; no posting performed

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.