Daily Recommend

v1.0.0

每日书籍、诗词、电影等优质内容推荐技能。通过不定时推送,帮助用户高频触达有价值的文学与艺术内容,培养日常阅读与审美习惯。触发条件:Cron 定时任务(三个随机时间窗口);用户主动说"推荐一首诗"等。

0· 134·0 current·0 all-time

Install

OpenClaw Prompt Flow

Install with OpenClaw

Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for billxfan/daily-recommend.

Previewing Install & Setup.
Prompt PreviewInstall & Setup
Install the skill "Daily Recommend" (billxfan/daily-recommend) from ClawHub.
Skill page: https://clawhub.ai/billxfan/daily-recommend
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.

Command Line

CLI Commands

Use the direct CLI path if you want to install manually and keep every step visible.

OpenClaw CLI

Bare skill slug

openclaw skills install daily-recommend

ClawHub CLI

Package manager switcher

npx clawhub@latest install daily-recommend
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
high confidence
Purpose & Capability
Name/description (daily recommendations, three random cron windows, manual triggers) align with the runtime instructions. The skill only requires scheduling, formatting recommendations, and recording simple preference data — nothing in the metadata asks for unrelated capabilities or secrets.
Instruction Scope
Instructions are narrowly scoped to generating one formatted recommendation per push, tracking user feedback into a workspace file, and deferring when recent conversation activity exists. The only potential ambiguity: it lists external content sources (Douban, Weixin 阅读, Zhihu, etc.) but does not specify how to fetch content (APIs, scraping, or human-curated prompts). That can affect behavior but is not inherently incoherent — platform/network policies determine allowed access.
Install Mechanism
Instruction-only skill with no install spec, no code files, and no downloads. This is the lowest-risk model and consistent with the stated functionality.
Credentials
The skill requests no environment variables or external credentials, which is proportional for simple recommendations. One small mismatch to be aware of: the SKILL.md mentions Feishu-style Markdown formatting and multiple third-party sources; if the skill later attempts to push to an external service (Feishu, Douban API, WeChat) it would need credentials that are not declared here. Also it writes a preference file to [WORKSPACE]/memory/recommend-preferences.md — ordinary for agent memory but users should be aware preferences are persisted to workspace storage.
Persistence & Privilege
always:false (not force-included) and autonomous invocation allowed (normal). The skill persists only its own small preference file under the agent workspace and does not request system-wide configuration or modify other skills. No elevated privileges are requested.
Assessment
This skill appears coherent and low-risk, but check a few practical points before installing: 1) Confirm where and how it will fetch content from external sites (APIs vs. scraping) and whether that requires credentials or violates terms of service. 2) Verify the delivery mechanism: if you want pushes routed through Feishu or other messaging platforms, those will need separate credentials not declared here. 3) Be comfortable with a small preference file being written to the agent workspace ([WORKSPACE]/memory/recommend-preferences.md) and review its retention/visibility. 4) Ensure the scheduling parameters (cron windows) and the 'check recent activity' behavior match your expectations and privacy preferences. If those items are acceptable or clarified, the skill is consistent with its stated purpose.

Like a lobster shell, security has layers — review code before you run it.

latestvk978tm57nfvxqv8avjvf9yy7bh83g33dliteraturevk978tm57nfvxqv8avjvf9yy7bh83g33dmoviesvk978tm57nfvxqv8avjvf9yy7bh83g33dpoetryvk978tm57nfvxqv8avjvf9yy7bh83g33drecommendationvk978tm57nfvxqv8avjvf9yy7bh83g33d
134downloads
0stars
1versions
Updated 1mo ago
v1.0.0
MIT-0

每日推荐技能 · 文学与艺术高频触达

核心定位

每天在三个随机时间窗口内,向用户推送一条诗词/名句/书籍/电影推荐。不是批量轰炸,是细水长流。

通过反馈机制持续学习用户偏好,逐步形成个性化的推荐画像,让推荐越来越精准。

核心原则

  1. 宁缺毋滥 — 每次只推一条,内容必须值得记住
  2. 有真情实感 — 推荐理由不罗列数据,要写出情感形状,让用户自己对号入座
  3. 多源并行 — 不依赖单一语料库,初始阶段多风格尝试,逐步聚焦
  4. 反馈驱动 — 根据用户反应调整偏好权重,像推荐系统一样持续学习

推送时间规则

每天在以下三个窗口内各随机选一个时间点触发:

窗口时间段说明
上午通过 cron 任务参数传入工作开始不久,适合文学性内容
下午通过 cron 任务参数传入午后疲劳,适合影视/轻松内容
晚上通过 cron 任务参数传入适合沉思性内容(诗词、名句、书籍)
  • 深夜(22:00–08:00)不推送
  • 触发前检查:若最近5分钟内有活跃对话,顺延30分钟

内容类型(每日轮换)

每次从以下类型中随机选一种:

  • 诗词:唐诗、宋词、元曲等,附情感解读
  • 名句摘录:文学作品或电影中的经典台词/段落
  • 书籍:小说、散文、传记等
  • 电影/纪录片:有叙事深度或美学价值的作品

推荐格式

每条推荐包含三个部分,使用 Feishu 支持的 Markdown 富文本:

【标题行】加粗标题 正文引用用代码块包裹,突出但不干扰阅读 为什么值得记住: 加粗标签 解读内容分点或分行,简洁有力 标签用 #标签 形式附在末尾


诗词推荐示例:

**《题都城南庄》· 崔护 · 唐**

> 去年今日此门中,人面桃花相映红。
> 人面不知何处去,桃花依旧笑春风。

**为什么值得记住:**
"人面不知何处去"写的是无常——某个普通的日子,熟悉的东西忽然不在了。
但诗人没有停在感伤里。"依旧笑春风",桃花还在好好开着——是一种"我知道你走了,但我还在好好生活"的态度。
不煽情,但就是会让人愣一下。

#唐诗 #无常 #错过 #桃花

名句推荐示例:

**✦《活着》· 余华 · 1993**

> "人是为了活着本身而活着,而不是为了活着之外的任何事物而活着。"

**为什么值得记住:**
这句话反直觉。我们习惯给活着找理由——为了家人、为了事业、为了意义。
余华说,不,活着的意义就是活着本身。
读进去之后会发现,这句话不是虚无,而是一种很深的释然。

#生命 #意义 #活着

书籍推荐示例:

**📚《一句顶一万句》· 刘震云 · 2009**

为什么值得记住:
刘震云写了那么多人——卖豆腐的、剃头的、染布的——每个人都活在孤独里。
不是没人说话,而是**找不到说得着的人**。
书里有一句:"世上的人遍地都是,说得着的人千里难寻。"
读到的时候像被人突然看穿了什么。

#孤独 #关系 #刘震云 #中国文学

格式规范:

  • 标题:加粗
  • 引用原文:代码块包裹(``
  • 关键句/关键词:加粗
  • 解读:普通文本,每句一行
  • 标签:前空一行,末尾用 #标签 形式

偏好追踪

每次推送后,[USER] 的反应会被记录到偏好文件中:

  • 喜欢(标记/回复正面) → 增加该类型/风格/作者的权重
  • 无感/忽略 → 降低权重
  • 明确不喜欢 → 记录关键词(情感虚假的、语言矫情的、说教的),后续规避

偏好文件路径:[WORKSPACE]/memory/recommend-preferences.md

触发方式

定时触发(Cron)

每天三个时间窗口各触发一次,使用 agentTurn 模式确保推送送达。

手动触发

用户说"来首诗"、"推荐本书"、"最近有什么电影"等,直接推荐一条。

信息来源

初期不锁定单一来源,多风格并行尝试:

  • 诗词:唐诗宋词经典库 + 豆瓣诗歌条目 + 各类诗词赏析
  • 名句:文学作品、电影台词、散文段落
  • 书籍:豆瓣读书、微信读书、知乎阅读推荐
  • 电影/纪录片:豆瓣电影、知乎影视高分榜

注意事项

  • 每条推荐只推一个内容,不堆砌
  • 推荐理由不写官方介绍词("情节跌宕起伏、引人入胜"之类)
  • 不确定用户偏好时,优先推诗词和名句(文学性最强、最容易触发共鸣)
  • 推荐后不做额外追问,等用户自然反馈
  • 遇到报错:主动解释原因,说明下一步处理

Comments

Loading comments...