X Ai Articles

# x-ai-articles An OpenClaw skill for finding **high-quality AI long-form content on X (Twitter)** and turning it into a curated daily digest. It is designed for people who want signal over noise: threads, deep dives, tutorials, breakdowns, launch writeups, and thoughtful analysis — not memes, one-line hot takes, or low-effort engagement bait. ## What this skill does - Searches X for AI-related long-form posts - Focuses on **threads, articles, guides, breakdowns, and deep dives** - Filters for **substantive content**, not just high-like posts - Ranks results by engagement and usefulness - Formats the output into a clean daily digest - Works well for **manual use** or **scheduled cron delivery** ## Typical use cases - “Show me today’s best AI threads on X” - “帮我抓一下 X 上今天最值得看的 AI 长文” - “Give me a daily AI reading digest from Twitter” - “Find the top AI deep dives from the last 24 hours” ## Requirements - [OpenClaw](https://github.com/openclaw/openclaw) - OpenClaw browser access - An active X/Twitter login session in the OpenClaw browser (`profile=openclaw`) ## Trigger phrases This skill is meant to activate on requests like: - `/aix` - `X AI articles` - `Twitter AI threads` - `AI trending posts` - `top AI tweets today` - `AI长文` - `X上的AI文章` - `帮我抓一下X上的AI长文` ## Installation ### Option 1: Copy into your local skills folder ```bash cp -r x-ai-articles ~/.openclaw/skills/ ``` ### Option 2: Clone from GitHub ```bash git clone https://github.com/clovericbot/x-ai-articles.git cp -r x-ai-articles ~/.openclaw/skills/ ``` ## How it works The skill uses the OpenClaw browser to search X with a query pattern like: ```text (AI OR LLM OR AGI OR Claude OR GPT OR OpenAI OR Anthropic OR "coding agent" OR "AI agent") (thread OR 🧵 OR article OR "deep dive" OR "here's what" OR "lessons" OR "breakdown" OR "tutorial" OR "guide" OR "wrote about" OR "published") min_faves:200 since:{yesterday} until:{today} ``` Then it: 1. opens X search results in the **Top** tab 2. extracts candidate posts from the page snapshot 3. filters for long-form / educational / analytical content 4. removes low-signal items 5. formats the best results into a ranked digest ## Output format Typical output looks like: ```md ## 📖 X 热门 AI 长文 | 2026-03-17 **1️⃣ Claude Code skills deep dive** @handle | ❤️ 8.8K | 🔖 1.2K | 👁️ 420K A practical breakdown of how Claude Code skills are structured, where they help, and how to build reusable coding workflows. https://x.com/handle/status/1234567890 ``` ## Quality filter The skill tries to keep only posts like these: ### Keep - Thread / 🧵 with educational value - Deep dives and technical breakdowns - Tutorials and practical guides - Launch posts with real explanation - Essays and thoughtful analysis ### Skip - One-line opinions - Memes / jokes / bait - Empty hype posts - Pure headlines without substance - Ads / promoted fluff ## Search tuning You can tune the search by changing: - **keywords**: add terms like `RAG`, `fine-tuning`, `agents`, `computer use` - **min_faves**: lower for niche topics, higher for mainstream quality control - **time range**: daily, weekly, or custom date ranges - **language filters**: e.g. `lang:en` ## Helper script The repo includes a small helper for generating search URLs: ```bash # Generate search URL for yesterday python3 scripts/build_search_url.py # Specific date python3 scripts/build_search_url.py 2026-03-17 # Date range with custom threshold python3 scripts/build_search_url.py 2026-03-15 2026-03-18 --min-faves=500 ``` ## Scheduling This skill also works well as a scheduled OpenClaw task. Example pattern: - **9:00 AM**: morning digest - **6:00 PM**: evening digest - target output: **5–10 strong items per run** ## Current repo status - GitHub: https://github.com/clovericbot/x-ai-articles - Primary use case: OpenClaw personal daily digest workflow - Can be adapted into a more general open-source X research / curation skill ## Roadmap ideas Possible future improvements: - deduplication across runs - JSON / Markdown export - archive mode for historical tracking - topic-specific presets - stronger scoring model for quality vs engagement ## License MIT