twitter

Read-only Twitter/X data lookup for tweets, users, followers, and engagement metrics. If the user has connected Twitter via Composio (OAuth), use Composio for account actions on their own profile: Rationale: Composio uses the user's own OAuth-connected Twitter account , so it can post, delete, and access self-endpoints. This skill uses a separate API key that is read-only and cannot act on the user's behalf. ⛔ CALL AT MOST 3 TWITTER TOOLS PER RESPONSE. STOP AFTER 3 CALLS. After each tool call, check: "Do I have enough data to answer?" If yes → STOP AND REPLY. ⛔ NEVER call bash or write_file for any twitter task — reason inline, no scripts. ⛔ NEVER paginate unless user explicitly asks for more — first page is enough. ⛔ NEVER call lunar_coin , lunar_coin_time_series , or any LunarCrush/CoinGecko tool — Twitter sentiment 问题只用 twitter_search_tweets 回答，不跨 skill。 ⛔ NEVER call coin_price , cg_trending , cg_coins_markets — 价格数据超出 Twitter skill 范围。 ⛔ NEVER use web_fetch for x.com or twitter.com URLs — Twitter blocks scraping, you'll only get a login wall. ✅ ALWAYS extract the tweet ID from the URL and use twitter_get_tweets . Example: User sends → Extract ID: 2042670029548794219 → Call: twitter_get_tweets(tweet_ids=["2042670029548794219"]) → Never: web_fetch(" Q: 找 3 个关于 BTC ETF 的高赞推文，只要 ID 和点赞数 → PLAN: 1 call twitter_search_tweets("BTC ETF min_faves:100") → pick top 3 from results →...