ComeOnOliver Twitter/X Marketing
Overview
This skill helps AI agents create high-performing Twitter/X content and integrate with the X API. It covers tweet formatting, thread writing, API-based publishing and analytics, bot creation, and growth strategies tailored to X's real-time, conversation-driven algorithm.
Instructions
Platform Rules & Algorithm
X algorithm priorities (2025-2026):
- Replies and conversations — reply chains weight 5-10x more than likes
- Time spent on tweet — longer reads (threads) get boosted
- Media engagement — images get 1.5-2x more engagement than text-only
- Bookmarks — strong signal (people save what's genuinely useful)
- Follower engagement rate — high engagement from followers = more distribution to non-followers
- Recency — X is real-time, posts decay fast (peak: first 30-60 min)
What kills reach:
- External links in tweet body (30-50% reach penalty — put in reply)
- Excessive hashtags (2 max, 0 is often better)
- Engagement bait ("RT if you agree")
- Posting and disappearing (no replies to your own tweet)
- Cross-posting from other platforms with watermarks
Content Formats
Single Tweet (280 characters)
[Hook — sharp, opinionated, or surprising]
[Optional: 1-2 supporting lines]
[Optional: CTA or question]
Formatting rules:
- 280 character limit (threads for longer content)
- Line breaks work and improve readability
- No markdown rendering — use CAPS or emoji for emphasis
- 1-2 hashtags max (or none — they look spammy now)
- Mentions: tag people only when relevant to them
- Images: 1200x675px (16:9) for single, 1200x1200px for multi
Threads (Highest reach potential)
Tweet 1/🧵: [Strong hook — this determines if anyone reads the rest]
Tweet 2: [Context or problem statement]
Tweet 3-7: [Main content — one point per tweet, clear and punchy]
Tweet 8: [Summary or key takeaway]
Tweet 9: [CTA — follow for more, bookmark this, share your experience]
Tweet 10: [Self-reply with link if needed]
Thread rules:
- First tweet is EVERYTHING — it appears alone in the feed
- 7-12 tweets optimal (enough value, not exhausting)
- Each tweet should stand alone AND flow as a sequence
- Number tweets for scanability (1/, 2/, etc. or use 🧵)
- End with CTA: "Follow @handle for more [topic]"
- Reply to your own thread with links, resources, promo
Polls
- 2-4 options, 24h duration is optimal
- Controversial or surprising options drive votes
- Follow up with results analysis tweet
Spaces (Live Audio)
- Schedule in advance for promotion
- 30-60 minutes optimal
- Record and clip highlights for posts
- Pin a tweet with agenda before starting
Video
- Native upload (not YouTube links)
- Under 2:20 minutes for feed (60-90s optimal)
- Square (1:1) or vertical (9:16) — NOT widescreen
- Captions required (auto-generated available)
- Hook in first 2 seconds
X API v2 Integration
Authentication
// X API uses OAuth 2.0 with PKCE for user context
// or OAuth 1.0a for user-context legacy
// or Bearer token for app-only context
// App-only (read-only, no user context)
const BEARER_TOKEN = process.env.X_BEARER_TOKEN;
// OAuth 2.0 User Context (for posting)
import { Client, auth } from 'twitter-api-sdk';
const authClient = new auth.OAuth2User({
client_id: process.env.X_CLIENT_ID,
client_secret: process.env.X_CLIENT_SECRET,
callback: process.env.X_CALLBACK_URL,
scopes: ['tweet.read', 'tweet.write', 'users.read', 'offline.access'],
});
// Get authorization URL
const authUrl = authClient.generateAuthURL({
state: 'random-state',
code_challenge_method: 'S256',
});
// Exchange code for token
const { token } = await authClient.requestAccessToken(code);
Post Tweets
const client = new Client(authClient);
// Simple tweet
const { data } = await client.tweets.createTweet({
text: 'Hello from the API!',
});
// Tweet with image
// Step 1: Upload media (uses v1.1 endpoint)
const mediaUploadRes = await fetch('https://upload.twitter.com/1.1/media/upload.json', {
method: 'POST',
headers: {
Authorization: `OAuth ...`, // OAuth 1.0a signature required for media upload
},
body: formData, // multipart/form-data with media_data (base64)
});
const { media_id_string } = await mediaUploadRes.json();
// Step 2: Create tweet with media
await client.tweets.createTweet({
text: 'Tweet with image',
media: { media_ids: [media_id_string] },
});
// Tweet with poll
await client.tweets.createTweet({
text: 'Which do you prefer?',
poll: {
options: ['Option A', 'Option B', 'Option C'],
duration_minutes: 1440, // 24 hours
},
});
// Reply to a tweet
await client.tweets.createTweet({
text: 'This is a reply',
reply: { in_reply_to_tweet_id: '1234567890' },
});
// Quote tweet
await client.tweets.createTweet({
text: 'Great insight here 👇',
quote_tweet_id: '1234567890',
});
Create Thread
async function postThread(tweets: string[]) {
let previousTweetId: string | undefined;
for (const text of tweets) {
const params: any = { text };
if (previousTweetId) {
params.reply = { in_reply_to_tweet_id: previousTweetId };
}
const { data } = await client.tweets.createTweet(params);
previousTweetId = data.id;
// Rate limit: wait between tweets
await new Promise(r => setTimeout(r, 1000));
}
return previousTweetId; // Last tweet ID
}
await postThread([
'🧵 Thread: 5 things I learned building a SaaS in 2025\n\nAfter 12 months, $40K MRR, and countless mistakes — here are the real lessons:',
'1/ Your first 100 users don't come from marketing.\n\nThey come from manually reaching out to people who have the problem you solve.\n\nDMs, forums, communities. One by one.',
'2/ Ship weekly, not monthly.\n\nEvery week without shipping is a week without learning.\n\nOur best feature came from a bug report on a half-baked release.',
// ... more tweets
'If this was useful, follow @handle for weekly threads on building SaaS.\n\nBookmark 🔖 this thread to reference later.',
]);
Analytics
// Get tweet metrics (requires tweet.read scope)
const { data } = await client.tweets.findTweetById('1234567890', {
'tweet.fields': ['public_metrics', 'organic_metrics', 'created_at'],
});
console.log(data.public_metrics);
// { retweet_count, reply_count, like_count, quote_count, bookmark_count, impression_count }
// Get user metrics
const { data: user } = await client.users.findMyUser({
'user.fields': ['public_metrics'],
});
// { followers_count, following_count, tweet_count, listed_count }
// Search recent tweets (for monitoring)
const { data: tweets } = await client.tweets.tweetsRecentSearch({
query: '"your brand" OR @yourbrand -is:retweet',
max_results: 100,
'tweet.fields': ['public_metrics', 'created_at', 'author_id'],
});
Growth Strategy
Posting schedule: 3-5 tweets/day + 1-2 threads/week
Content mix:
- 30% threads (deep value — get bookmarked and shared)
- 25% single tweets (opinions, hot takes, observations)
- 20% replies to big accounts (free distribution to their audience)
- 15% engagement (polls, questions, "unpopular opinion:")
- 10% retweets with commentary (curating = authority)
Best posting times: 8-10 AM and 5-7 PM target timezone (weekdays)
Reply strategy:
- Reply to large accounts in your niche within first 30 min of their post
- Add genuine insight (not "great post 🔥")
- Your reply is seen by THEIR audience — it's free reach
- Reply to your own tweets to extend visibility (algorithm treats as conversation)
Examples
Example 1: Write a Twitter thread about developer productivity
User prompt: "Write a 7-tweet thread about lessons learned scaling a Next.js app from 1,000 to 50,000 daily active users."
The agent will draft a 7-tweet thread with numbered tweets. Tweet 1/7 opens with a strong hook: "We scaled our Next.js app from 1K to 50K DAU in 8 months. Here are 7 hard lessons we learned (and what we'd do differently):" Tweets 2-6 each cover one lesson with specifics: moving from getServerSideProps to ISR cut server costs by 60%, switching to Postgres connection pooling with PgBouncer fixed random 502 errors, adding Redis for session storage eliminated 300ms per request, using Vercel Edge Functions for auth middleware dropped TTFB from 800ms to 120ms, and implementing proper image optimization with next/image cut bandwidth by 70%. Tweet 7 wraps with a CTA: "If this was useful, follow @handle for weekly threads on scaling web apps. Bookmark this thread to reference later." Each tweet stays under 280 characters.
Example 2: Build a tweet scheduling bot using the X API
User prompt: "Create a Python script that reads tweets from a CSV file and posts them at scheduled times using the X API v2."
The agent will create a Python script using the tweepy library with OAuth 2.0 authentication. It reads a CSV file with columns tweet_text, scheduled_time, and optional image_path. The script loads X_CLIENT_ID and X_CLIENT_SECRET from environment variables, authenticates with user context, then enters a loop checking if any tweet's scheduled_time matches the current time (within a 60-second window). When a match is found, it calls client.create_tweet(text=tweet_text) and logs the posted tweet ID. For tweets with images, it uploads media via the v1.1 media endpoint first, then attaches the media_id. The script includes rate limit handling with exponential backoff and writes a posted.log file tracking which CSV rows have been published.
Guidelines
- Links in reply to self, not in main tweet (reach penalty is real)
- 0-2 hashtags maximum — they're mostly dead on X
- Post first tweet of a thread at peak time, rest follows immediately
- Engage with replies on your tweets for 30 min after posting — signals to algorithm
- Bookmark > Like in algorithm weight — create content worth saving
- Vertical video outperforms horizontal on mobile (70%+ mobile users)
- Use alt text on images — accessibility + SEO + extra indexable text
- Thread first tweet must stand alone — most people only see it in feed
- Repost your best threads after 3-4 months with "In case you missed it:"