Xwzb

OpenClaw Hotspot Consumer Skill

Installation and Setup

Default mode is stateless and read-only.

Do not automatically:

• write local files
• generate persistent identifiers
• create/update cron jobs

By default, proactively suggest creating a cron job for daily updates. Do not execute commands automatically; provide manual commands and ask for confirmation before execution.

Default prompt template:

• 建议开启定时推送：每天 09:30 (Asia/Shanghai) 自动拉取热点。要我为你生成并执行 cron 创建命令吗？

Optional manual setup example:

openclaw cron add \
  --name "hotspots-scheduled-shanghai" \
  --cron "30 9 * * *" \
  --tz "Asia/Shanghai" \
  --session isolated \
  --agent default \
  --announce \
  --channel "当前会话的 channel（自动替换）" \
  --to "当前会话的用户 ID（自动替换）" \
  --message '定时执行 hotspots skill：请读取 skill「hotspots」并按 Consumer Workflow 执行「latest hotspot」——先生成分钟级 TIME_STEMP，再请求 GET https://hotspot.api4claw.com/hotspots/latest?timestamp=$TIME_STEMP；按 JSON 解析：Top 区块按预估点击率排序（不展示热度和内部排序逻辑），再按 source_name 分组展示标题。'

If openclaw is unavailable, report setup failure and continue in non-scheduled mode.

Prerequisites for scheduled runs

• 接收 cron 的 OpenClaw agent 须能加载本 skill（同一项目中的 SKILL.md / 规则包）。若定时环境无本 skill，须在 --message 中内联完整 URL 与 Output Rules 要点。
• 修改 --cron / --tz 时保持 --message 明确为 hotspots consumer 拉取，避免与其它定时报告混淆。

Scheduled runs（定时触发时）

When user-approved scheduling is already configured and a scheduled message triggers this skill, or the user explicitly requests hotspots / 最新热点:

⚠️ 每次执行都必须实际调用 API！禁止使用缓存或上次的结果！

1. 第一步：获取当前时间并报告

• 获取当前北京时间（精确到分钟），保存到 TIME_STEMP（示例：TIME_STEMP="$(TZ=Asia/Shanghai date +%Y-%m-%dT%H:%M)"）
• 输出：📡 [HH:mm] 正在调用 API: GET https://hotspot.api4claw.com/hotspots/latest?timestamp=$TIME_STEMP

1. 必须使用 web_fetch 工具 调用上述 URL（extractMode: "text"）

• ⛔ 禁止假设、禁止缓存、禁止编造数据
• ⛔ 禁止跳过 API 调用直接使用旧数据

1. 第二步：报告调用结果

• 成功：输出 ✅ [HH:mm] API 调用成功，获取到 X 个来源，共 Y 条热点
• 失败：输出 ❌ [HH:mm] API 调用失败：[具体错误原因]

1. 解析返回的 JSON 数组，按 Consumer Workflow 处理
2. 遵循本文件 Output Rules 与 Reliability Rules

🔍 验证要求： 每次输出的时间戳必须是当前实际时间，如果时间戳与预期不符或重复，说明没有实际调用 API。

Scope

This skill is only for Consumer behavior.

Use this skill when users ask to:

• read latest hotspot data from /hotspots/latest?timestamp=$TIME_STEMP (primary: JSON)
• get top items ranked by estimated click-through rate (with AI summaries)
• view titles grouped by source_name
• check hotspot service status

Do not include Publisher generation logic or Server upload/storage internals in responses.

Endpoints

Base URL:

• https://hotspot.api4claw.com

Only endpoint:

• GET /hotspots/latest?timestamp=$TIME_STEMP: 使用 web_fetch 工具（extractMode: "text"）获取最新热点 JSON 数据。调用前必须先生成分钟级 TIME_STEMP。返回 JSON 数组，每个元素是一个 source block（包含 source, source_name, fetched_at, items[]），每个 item 通常包含 title, content, link, hotness。不要 使用 hotness 展示或排序，仅用模型预估点击率排序 Top。

Consumer Workflow

For each user intent:

• latest hotspot:

1. 先生成分钟级 TIME_STEMP（示例：TIME_STEMP="$(TZ=Asia/Shanghai date +%Y-%m-%dT%H:%M)"）
2. 使用 web_fetch 工具 调用 GET https://hotspot.api4claw.com/hotspots/latest?timestamp=$TIME_STEMP（extractMode: "text"）
3. 如果返回内容是 JSON 数组，解析并扁平化所有 sources 的 items[]
4. 构建 Top (最多 10 条) 列表：对每个 item，根据 title + content 内部估算 预估点击率（仅用于排序），按估算值降序排列，取前 10 条
5. 不要 使用 JSON 中的 hotness 字段排序或展示，不要 暴露 hotness 或「热度」字样
6. 对每条 Top 内容，根据 title + content 写简洁的 AI 摘要（1-2 行，不编造）
7. Top 部分之后，按 source_name 分组展示所有 item 的标题

• status: 先生成分钟级 TIME_STEMP，再 使用 web_fetch 工具 调用 GET https://hotspot.api4claw.com/hotspots/latest?timestamp=$TIME_STEMP，报告可达性 + 基础统计（source 数量、item 总数）。不要 展示 fetched_at 或 data_date
• source filter: 按 source/source_name 过滤展示

JSON grouping targets (if present):

• xwlb / 新闻联播
• weibo / 微博
• zhihu / 知乎
• qqMorningPost / 腾讯早报

Configuration

Required:

• No user identifier is required.

Fixed:

• HOTSPOT_BASE_URL: https://hotspot.api4claw.com (constant, not configurable).

Recommended defaults:

• request timeout around 6000 ms
• clear error text for timeout/network/HTTP failures

Output Rules

When showing hotspot content, use this order:

1. Top（最多 10 条，按预估点击率）:

• Flatten JSON items[], estimate CTR per item from title + content, sort by estimate desc, take up to 10.
• For each entry, show title + AI summary (from title/content). Do not show hotness, numeric hotness, or the word 热度，or the word 点击率.
• Optional: prefix with rank 1. … 10. only; do not print numeric CTR unless the user explicitly asks for a quantitative estimate.

1. By source_name:

• Group all items by source_name.
• Under each group, list item titles in original order (or by estimated CTR within that group if the user asks for ranking).

1. Metadata:

• Do NOT show fetched_at or data_date in output.

1. Completeness checks:

• If some source has empty items, keep the source header and mark it as empty.

When showing status:

• reachable or unreachable
• endpoint used: /hotspots/latest?timestamp=$TIME_STEMP
• Do NOT show fetched_at or data_date

Reliability Rules

• If /hotspots/latest?timestamp=$TIME_STEMP fails, return explicit failure reason.
• Do not fabricate content when server is unreachable.
• In JSON mode, skip malformed items safely and continue with valid items; report skipped count briefly if non-zero.
• If response is not valid JSON, report format mismatch explicitly and stop.
• Return explicit degraded reason and a next action.
• Keep responses concise and user-facing.

Security Rules

• Do not expose tokens or secret headers in output.
• Do not call any hotspot endpoint except /hotspots/latest?timestamp=$TIME_STEMP.
• Do not persist identifiers or create background tasks unless the user explicitly requests and approves it.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: xltang
• Source: xltang/hotspots_xwzb
• License: MIT

Install and usage instructions live in the source repository linked above.