{"manifest":{"name":"X Content Optimizer","version":"1.0.0","description":"Write and optimize X/Twitter content using exact engagement weights from Twitter's open-source algorithm, real reach killers, and high-value behaviors that compound distribution.","tags":["twitter","x","social-media","content","algorithm"],"standard":"agentskills.io","standard_version":"1.0","content_checksum":"7dca4da67b8ed8cb317e6a2a53b0fadea5559ea8791f545ff42966ba9d5e7b7d","bundle_checksum":null,"metadata":{},"files":[]},"files":{"SKILL.md":"# X Content Optimizer\n\nWrite and optimize X (Twitter) content using the exact engagement weights from Twitter's open-source algorithm, the real reach killers, and the high-value behaviors that compound into distribution.\n\n## Overview\n\nTwitter's recommendation algorithm has been open-sourced at `twitter/the-algorithm`. The engagement weights are public. This skill encodes those weights and the known reach dynamics into a practical writing framework for anyone creating content on X.\n\n**Based on the Twitter open-source algorithm weights and community X algorithm research.**\n\n## The Algorithm Weights\n\nThese are the engagement event weights that determine how the algorithm scores a tweet's quality:\n\n| Engagement | Weight | Notes |\n|---|---|---|\n| Author replies to your reply | 75× | Highest possible signal |\n| Direct reply to tweet | 27× | (heavy engagement) |\n| Direct reply from non-follower | 13.5× | Broader reach signal |\n| Retweet | 20× | High share signal |\n| Profile click through to profile | 12× | Shows genuine interest |\n| Link click | 11× | (link inside app) |\n| Bookmark | 10× | High-intent save |\n| Like | 0.5× | Baseline engagement |\n| Video 50% completion | ~0.005× | Very low weight |\n\n**Key insight:** A single reply chain where the author responds is worth 150× a like. Build content that generates conversation, not just likes.\n\n## The Reach Killers\n\nThese actions actively suppress a tweet's distribution:\n\n### External Links — 50–90% Reach Reduction\nTweets with links to external sites (any domain not Twitter/X) are suppressed in the For You feed. The algorithm deprioritizes content that takes users off-platform.\n\n**Strategy:** Put the link in a reply. Post the main tweet link-free, then reply to yourself with the URL. The main tweet distributes; the reply captures clicks.\n\n### Repetitive Content / Engagement Bait\nThe algorithm flags patterns like \"RT to win,\" \"like if you agree,\" explicit asks for engagement. These trigger the \"engagement manipulation\" classifier.\n\n**Strategy:** Never ask for engagement directly. Create content worth engaging with.\n\n### A Single Report Can Destroy Reach\nOne report from a user can dramatically reduce a tweet's reach until review. This asymmetric downside means controversial framing is a distribution risk, not just a sentiment risk.\n\n### Rapid-Fire Posting Bursts\nPosting multiple tweets in a short window signals spammy behavior. The algorithm normalizes for burst activity.\n\n**Strategy:** Space posts with at least 30-60 minutes between them during a campaign window.\n\n## High-Value Behaviors\n\n### Replies-to-Replies Are 75× — Build Conversations\n\nThe highest weight event is the original author replying to someone who replied to them. This means:\n- When someone engages with your tweet, reply back quickly\n- The algorithm interprets author-responds-to-reply as a signal that the content sparked real conversation\n- A thread where you respond to top replies compounds faster than a thread you write alone\n\n**Workflow:** Post → monitor for the first 30 minutes → reply to every substantive reply immediately. This initial engagement burst is the most important window.\n\n### Profile Clicks Are 12× — Make the Profile Worth Clicking\n\nDwell time on a tweet followed by a profile click is a very strong quality signal. This means your bio, pinned tweet, and recent posts are integral parts of every tweet's distribution.\n\n**Checklist:**\n- Bio states specifically what you do and why people should follow\n- Pinned tweet is your best recent post (update monthly)\n- Profile picture is recognizable at 40px thumbnail size\n- Header image reinforces the bio message\n\n### Video Gets ~10× More Distribution Than Text-Only\n\nVideo tweets receive significantly more For You feed placement than text-only tweets. The algorithm favors content that keeps users on platform.\n\n**Practical note:** Dwell time threshold is approximately 15 seconds. A video under 15 seconds that users don't complete scores poorly. A video over 30 seconds with 50%+ completion scores well.\n\n### The First 30 Minutes Are Decisive\n\nThe algorithm uses early engagement velocity to determine wider distribution. A tweet that gets 20 replies in 30 minutes gets pushed broadly. A tweet that gets 2 replies in 30 minutes enters a low-visibility pool.\n\n**Strategy:** Post at your highest-engagement time window. Use analytics to find when your followers are most active (typically 8–10am, 12–1pm, and 8–10pm in their local time zone).\n\n## Tweet Structure for Maximum Weight\n\n### Hook optimization\nThe first sentence (visible without \"Read more\") must generate curiosity or state a specific claim. Vague openings lose the scroll-stop. Specific claims (with numbers, names, or counterintuitive statements) perform 3–5× better than generic openings.\n\n**High weight:** \"The Twitter algorithm source code has 6 variables that determine your reach. None of them are follower count.\"\n\n**Low weight:** \"Here are some thoughts on the Twitter algorithm that I think you'll find interesting:\"\n\n### The Reply-Anchor Thread Structure\n\nFor threads with important information:\n1. Post the most engaging insight as the first standalone tweet (no \"1/N\")\n2. Reply to yourself with the thread content\n3. The first tweet distributes; thread readers who want more find it in replies\n\nThis format outperforms traditional numbered threads because the first tweet is self-contained and not marked as \"part 1 of N\" (which signals long-form commitment and reduces click-through).\n\n### Quote Tweet vs Retweet\n\n- Retweet: 20× weight, but credit goes to original author\n- Quote Tweet with substantial added commentary: ~27× (treated as a reply with retweet signal)\n- Quote Tweet with only \"interesting!\" type commentary: scores closer to retweet, less distribution\n\n**Rule:** Only quote tweet when you have something genuinely additive to say. Otherwise, retweet.\n\n## Pre-Post Checklist\n\n```\n[ ] No external links in main tweet (link in first reply if needed)\n[ ] No explicit engagement asks (\"RT if\", \"like if\", \"share with\")\n[ ] Hook sentence: specific claim or counterintuitive statement\n[ ] Post time: within your audience's peak engagement window\n[ ] Profile is current and bio states specific value proposition\n[ ] First 30 minutes: calendar a reply review window after posting\n[ ] For video: runtime > 30s, core content in first 15s\n[ ] For threads: first tweet is self-contained, not labeled \"1/N\"\n```\n\n## Measuring Success\n\nThe metrics that matter per the algorithm weights, ranked:\n\n1. **Reply rate** — the highest-weight signal per impression\n2. **Retweet rate** — second highest aggregate signal\n3. **Profile click rate** — indicates audience growth potential\n4. **Bookmark rate** — high-intent engagement (saves for later)\n5. **Like rate** — lowest weight, commonly over-optimized\n\nIf you're optimizing for reach, optimize for reply rate and retweet rate. Likes are a vanity metric relative to the algorithm.\n"}}