← 返回 Skills 市场

Hourly Knowledge

Name: Hourly Knowledge
Author: mrmiracle1

作者 MrMiracle1 · GitHub ↗ · v1.0.0 · MIT-0

cross-platform ⚠ suspicious

总下载

当前安装

版本数

在 OpenClaw 中安装

/install hourly-knowledge

功能描述

每小时自动为用户推送一条有趣、有用的小知识。当每小时定时任务触发时使用，负责搜索并生成知识内容。包含去重机制，避免短期内推送相同主题。支持多用户，开箱即用。

使用说明 (SKILL.md)

Hourly Knowledge

概述

每小时整点触发时，为用户推送一条有趣的小知识。支持主题去重，确保短期内容不重复。

核心设计

用户上下文：通过 inbound meta 自动获取当前用户的 chat_id 和 channel
配置分离：敏感信息（accountId）放在 config.json，不在代码中硬编码
多用户友好：不同用户使用同一份 skill 代码，互不影响
推送机制：通过 cron announce 机制自动推送，subagent 只返回纯文本

工作流程

从 config.json 读取 accountId
从会话上下文获取当前用户的 chat_id 和 channel
读取 memory/hourly-knowledge-history.json 获取最近推送的主题列表
用 web_search 搜索一个不在最近列表中的新主题
生成知识内容（150字以内，配 emoji）
将新主题写入历史记录
直接返回纯知识文本，由 cron announce 机制自动推送到用户

搜索策略

主题范围：人文、历史、科技、自然、生活、地理、艺术等
避开最近 10 条已推送过的主题
选择大多数人不知道但有趣、有用的知识点

输出格式

直接返回纯文本，例如：

📚 小知识：

🧠 为什么手指泡水后会变皱？

人的手指在水中泡久了会起皱，这是因为皮肤最外层的角质层吸水膨胀，而真皮层没有变化，于是表皮就产生了褶皱。这是神经系统在告诉你：物体湿滑，需要更好的抓握能力去抓东西！

注意事项

每次必须从历史记录中排除最近 10 条主题
不要解释"我是谁"或"这是什么任务"
不要调用 message 工具，只返回纯文本
直接返回内容，不回复 HEARTBEAT_OK
返回的文本就是用户收到的推送内容

配置文件

config.json 必须与 SKILL.md 同目录，内容示例：

{
  "accountId": "487efdd501be-im-bot"
}

安全使用建议

This skill appears to implement hourly knowledge pushes, but there are some implementation details you should check before installing: - Multi-user claim vs implementation: the code stores history at a single global path (/root/.openclaw/workspace/memory/hourly-knowledge-history.json). If you expect per-user de-duplication or privacy separation, this is not implemented; ask the author to use per-user keys/filenames or adjust the code to incorporate user identifiers. - Config sensitivity: config.json contains an accountId (not a secret token in the provided sample), but confirm whether that identifier is sensitive in your deployment and who can read the skill directory. - Inbound meta mismatch: SKILL.md says the agent will get chat_id/channel from inbound meta, but the included script does not reference inbound meta. Make sure the runtime agent logic that composes messages supplies per-user context correctly and that the skill will not accidentally use the global history for all users. - Message delivery and autonomy: the SKILL.md instructs the skill not to call the message tool and to return plain text, relying on a cron/announce mechanism to push messages. Confirm how your platform's cron announce works and what permissions are used to deliver messages. Test in a staging environment to confirm behavior. - Data persistence and access control: the script writes to /root/.openclaw/workspace; verify filesystem permissions so other processes or users cannot tamper with or read the history if that is a concern. Overall: the footprint is small and there are no external network downloads or credentials requested, but the cross-user history implementation is a real coherence issue. If you need strict per-user semantics, request or implement a fix before enabling this skill for multiple users.

能力评估

⚠ Purpose & Capability

The description and SKILL.md describe an hourly multi-user knowledge pusher with topic de-duplication and a config-held accountId. The included helper script aligns with tracking topics and reading a config file. However the skill claims '多用户互不影响' (multi-user isolation) but the implementation uses a single global history file at /root/.openclaw/workspace/memory/hourly-knowledge-history.json and a single skill config path, which does not provide per-user separation as claimed.

⚠ Instruction Scope

SKILL.md instructs the agent to read config.json, read/write a recent-topics history, obtain chat_id/channel from inbound meta, use the platform web_search tool, and return pure text (no message tool). The helper script handles reading/writing history and config, but it does not interact with inbound meta or any per-user context — that gap between the prose (use inbound meta) and the script (no such code) is an inconsistency. The instructions otherwise stay within the declared purpose and do not ask for unrelated files or credentials.

✓ Install Mechanism

No install spec is present (instruction-only plus a small helper script). No external downloads or package installs are requested. This is low-risk from an installation/third-party code distribution perspective.

ℹ Credentials

The skill requests no environment variables or external credentials. It uses a local config.json containing an accountId string (not a secret token). That's proportionate to the described functionality. Consider whether that accountId is sensitive in your environment — it is stored in the skill directory and read by the script.

ℹ Persistence & Privilege

The skill writes to workspace memory at an absolute path under /root/.openclaw/workspace/memory, which is consistent with storing its own history. It is not marked always:true and does not request extra system privileges. However, the single shared history file implies persistent, global state across users/contexts rather than isolated per-user state, which can cause cross-user leakage or incorrect de-duplication behavior.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install hourly-knowledge
安装完成后，直接呼叫该 Skill 的名称或使用 /hourly-knowledge 触发
根据 Skill 的参数说明提供必要输入，即可获得结构化输出

版本历史

v1.0.0

- Initial release: Automatically sends an interesting, useful fact to users every hour. - Supports multiple users with isolated context; chat_id and channel detected automatically. - Prevents duplicate topics by tracking recent 10 sent themes. - Facts are generated within 150 characters and include emojis. - Pushes messages as plain text; no extra formatting or explanations. - Configuration is separated; sensitive info (accountId) stored in config.json.

元数据

Slug hourly-knowledge

版本 1.0.0

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 1

常见问题