A Note-Based Workflow for Personalized Agent Recommendations
A workflow shared by a flomo Agent user asks a memory-enabled agent to select exactly three items each day to read, watch or listen to in Chinese. The first two should address the user’s current concerns based on recent conversations without repeatedly staying on one topic, while the third should broaden exposure to areas such as art, design, games, music, humanities, nature and practical disciplines. Each recommendation should include a title, a verified original link, a brief description and why it is worth the user’s attention. The workflow also calls for checking accessible recent recommendations and conversations, avoiding content the user has explicitly read, distinguishing source findings from personal interpretation, and disclosing access limitations. The author says they have used it in flomo Agent with good results and that it can be used by memory-enabled agents such as grokbot or dot.
The workflow offers a concrete method for turning notes and recent conversations into varied recommendations while requiring source verification and explicit limits on what the agent can establish. Its reported results come from the author’s personal use, and broader compatibility is asserted rather than demonstrated.
最近在 flomo Agent 一直用这个方法，把笔记当做推荐依据获取信息，效果还不错。 当然，在任意有记忆的 Agent 也都可以使用，比如 grokbot 或者 dot - 每天为我挑选恰好3个值得阅读、观看或聆听的具体内容，用中文推送。 围绕真正重要的问题选择，优先有新证据、新视角或能挑战既有判断的内容，避免泛泛的 AI 新闻汇总。 前2条结合我当前关心的问题，根据近期对话判断，不要长期固定在同一主题。 第3条专门拓宽视野：从艺术、设计、游戏、音乐及更广的人文、自然与实践领域，选择我可能未接触过的具体作品、人物、方法或现象。跨天保持很高的领域、媒介、时代、地域和思考方式多样性。不要总围绕 AI、产品、效率或知识管理挑相邻题材，也不要强行把每项都转化为产品启示。
优先陌生但有可理解入口、有趣且有实质启发的内容，不为冷门而冷门。 每条给出： * 标题 * 经过核验的原始链接 * 简短介绍：它具体讲了什么 * 为什么值得我看 明确区分来源本身的观点、研究结论与你自己的联想，不把个人经验当成实验证据，也不夸大研究能证明什么。 检查能够访问的近期推送和对话，避开重复与我明确读过的内容。无法证明从未接触时，只能说“可能陌生”，不能声称确定。 使用公开网络检索，并实际阅读原始来源验证内容。优先免费可访问的来源，披露付费墙、只读到摘要、未实际观看或收听等限制。涉及当前事实时核对时效，必要时标明具体日期。 保持简洁、有判断，不布置阅读任务。仅做阅读研究和向我推送，不联系第三方，不更改账户或订阅。