據 AlphaMoat 統計,截至 2026-09,Content Research 累計下載 4,059 次、獲 1 個星標,在全部 5,078 個 AI 技能中排名第 2,503 位,在「內容創作」分類(362 個)中排名第 175 位。
| 平台 | 累計安裝 | 累計下載 | 星標 | 評論 |
|---|---|---|---|---|
| ClawHub ↗ | 115 | 3669 | 1 | - |
| SkillHub ↗ | 113 | 4059 | 1 | — |
各平台的計數口徑不同,統計時間不一致。
先搜尋並篩選近期新聞,再按 Reddit、X/Twitter、Discord 或 LinkedIn 的格式生成摘要或帖子草稿;可採用不同內容角度或參考品牌語氣。釋出前應核實來源連結和內容準確性。 品牌配置檔案不要寫入秘密或賬號憑據。
資料釋出於 2026-10-29
以下是根據技能文件生成的示例,不代表實際執行結果。
2026-09 起統計方法調整:同名技能按作者分別統計,並覆蓋全部技能。圖中虛線為調整前的月份,與之後的資料不直接比較,也不計算跨越調整的增長。 該技能有同名的其他版本,調整前的資料歸屬當前下載量最大的這一個。
| 月份 | 累計下載 | 環比 | 星標 |
|---|---|---|---|
| 2026-09 | 4059 | — | 1 |
| 2026-08 | 3530 | +5.1% | 1 |
| 2026-07 | 3358 | +12.2% | 1 |
| 2026-06 | 2992 | +6.9% | 1 |
| 2026-05 | 2799 | — | 1 |
| 2026-03 | 1571 | — | 1 |
以下是作者釋出的原始檔案,按純文字顯示,本站未做修改或稽核;其中的連結不可點選,命令請核實後再執行。抓取於 2026-10-23。
--- name: content-research description: Research trending topics and generate platform-specific content. Triggers on "research [topic]", "what's new in [topic]", "content for [platform]", "create posts about [topic]". Supports Reddit, X/Twitter, Discord, LinkedIn with multiple content angles per platform. --- # Content Research Two-phase workflow: **Research** → **Create** ---
## Phase 1: Research **Triggers:** `research [topic]`, `what's new in [topic]` ### 1. Search Use web search to find recent news: ``` web_search(query="[topic] news", freshness="pw") ``` **Query patterns:** - News: `[topic] news` - Reddit: `site:reddit.com [topic]` - X/Twitter: `site:x.com [topic]` ### 2. Fetch Articles Extract article content: ``` web_fetch(url="[URL]", maxChars=8000) ``` ### 3. Filter - **7-day cutoff** — discard older content - Skip "what is X" explainers - Skip price predictions / TA - Prioritize: launches, partnerships, updates, drama, milestones ### 4. Present ```markdown ## [Topic] Research — [Date] 1. **[Headline]** - [Source] - [X days ago] [2-3 sentence summary] 2. **[Headline]** - [Source] - [X days ago] [2-3 sentence summary] [up to 5 items, newest first] ``` --- ## Phase 2: Content Creation **Triggers:** `create content for [platform]`, `#3 for reddit` ### Platform Formats #### Reddit - Hook title (no clickbait) - 2-4 conversational paragraphs - Include source link - End with discussion prompt **Angles:** 1. **News share** — Straightforward reporting 2. **Discussion** — "What do you think..." 3. **Analysis** — Your take on implications 4. **ELI5** — Simple explanation 5. **Contrarian** — Devil's advocate #### X/Twitter - Under 280 chars (or thread) - Hook first line - Line breaks for readability **Angles:** 1. **Breaking** — Just facts, urgency 2. **Hot take** — Engagement bait opinion 3. **Thread** — Multi-tweet breakdown 4. **Quote dunk** — React to announcement 5. **Meme** — Casual/funny #### Discord - Bullet lists (no tables) - Wrap links: `<https://...>` - Bold/CAPS for emphasis **Angles:** 1. **Alert** — One-liner + link 2. **Summary** — Key bullets 3. **Discussion** — Ask for reactions 4. **Thread** — Detailed breakdown 5. **Meme** — Community vibe #### LinkedIn - Professional tone - Lead with insight - 3-5 short paragraphs - End with question **Angles:** 1. **Industry insight** — What it means 2. **Lessons** — What we learn 3. **Prediction** — Where it's heading 4. **Career** — Professional implications 5. **Case study** — Deep dive --- ## Brand Voice (Optional) For branded content, create a `brand-config.md` file with your voice guidelines: ```markdown # Brand: [Name] ## Voice - [Tone descriptor] - [Communication style] ## Avoid - [Things not to say] ## Include - [Required elements] ``` When generating branded content, reference your brand config for consistency. --- ## Example Session ``` User: research defi Agent: [Returns 5 findings from past 7 days] User: 2 for reddit Agent: [5 Reddit angles for finding #2] User: angle 3 Agent: [Ready-to-post content] ```
識別並去掉文字里的 AI 寫作痕跡,讓文字讀起來更像人寫。
把文字里 AI 寫作的典型痕跡挑出來並改寫得更自然,同時保留原意的寫作編輯技能。
識別並去除 AI 生成文字痕跡的中文文字編輯技能。
wechat-search 檢索公眾號熱門文章,並按相關性、熱度和時效整理結果。
從公開短影片分享連結提取標題、簡介和口播文案,並給出改寫版本。
humanizer 為中英文文字提供更自然的表達和風格編輯指導。
寫網文章節並按檔案落盤,聊天裡只回摘要與指標。
把一組行銷方法模組收進同一個技能,按請求路由到對應模組取清單與可交付稿。
資料月份:2026-09 · 下載/星標/安裝數聚合自公開技能註冊庫(ClawHub、SkillHub),按月更新。口徑說明見資料方法。