daily-news-briefing

title	Daily News Briefing Workflow
emoji	📈⚔️💻
name	daily-news-briefing
description	Automates the generation and delivery of a daily news briefing covering finance, military, and technology sectors.
usage	Use this skill to generate and push a daily news briefing covering finance, military, and technology sectors.
input_schema	{"type":"object","properties":{"schedule_time":{"type":"string","description":"The time to send the briefing (default is '10:00 AM')."},"sections":{"type":"array","items":{"type":"string"},"description":"List of sections to include (default is ['finance', 'military', 'technology'])."}},"required":[]}

Daily News Briefing Workflow

This skill automates the process of generating and pushing a daily news briefing with real-time updates across multiple sectors.

Prerequisites

This user's environment uses Firecrawl API v2 (key: fc-5805f00749234abf8c75c682748a2633, endpoint: /v2/search) for supplemental search. Sina Feed API is the primary source for same-day news due to Firecrawl's unreliable date filtering. The productivity/news-briefing skill has additional Feishu push details.

For weather jobs: free wttr.in API requires no key.

News Sources & Working APIs

Finance & Tech — Sina Feed API (reliable, fast)

https://feed.mix.sina.com.cn/api/roll/get?pageid=153&lid={lid}&k=&num=15&page=1&r=0.5

Section	lid	Notes
Finance	2516	Returns ~50 items, timestamps in Unix epoch
Tech	2515	Returns ~50 items, timestamps in Unix epoch
General	2517	⚠️ Returns finance content — NOT reliable for military. Use browser for military.

Response shape: {"result": {"data": [{"title", "intro", "ctime" (Unix timestamp), "url"}]}}

Filtering for military content from lid=2517: keywords include 军, 武, 舰, 导, 战, 部队, 军事, 国防, 伊朗, 以色列, 俄罗斯, 乌克兰, 北约.

Military — Browser Scraping Required (API returns stale data)

⚠️ Critical: The Sina military API (lid=2514) returns cached data from 2023. Do NOT use it. Instead:

1. Use browser_navigate('https://mil.news.sina.com.cn/')
2. Use browser_snapshot() to extract the news list — headings and timestamps are in the page accessibility tree
3. Use browser_click(ref) to open individual articles
4. Extract article content via browser_console("JSON.stringify([...document.querySelectorAll('p')].map(p=>p.textContent))")

The military page renders dates as "今天 HH:MM" (today) or "6月9日 HH:MM" (recent).

Steps

1. Fetch Finance News

   # Use lid=2516, iterate items and convert ctime
   import subprocess, json
   from datetime import datetime
   result = subprocess.run(['curl', '-s', '--max-time', '15',
       'https://feed.mix.sina.com.cn/api/roll/get?pageid=153&lid=2516&k=&num=15&page=1&r=0.5',
       '-H', 'Referer: https://news.sina.com.cn/'],
       capture_output=True, text=True, timeout=20)
   data = json.loads(result.stdout)
   for i in data['result']['data']:
       ts = int(i['ctime'])
       dt = datetime.fromtimestamp(ts).strftime('%Y-%m-%d %H:%M')
       print(f"[{dt}] {i['title']} | {i['intro']} | {i['url']}")
   

2. Fetch Tech News — same API with lid=2515

3. Fetch Military News — browser only (see above)

4. Extract Article Detail (optional, for fuller summaries):

   # Sina article pages may use GB2312 encoding
   result = subprocess.run(['curl', '-s', '--max-time', '15', URL,
       '-H', 'User-Agent: Mozilla/5.0'], capture_output=True, text=True, timeout=20)
   text = result.stdout  # bytes — decode with errors='ignore'
   paras = re.findall(r'<p[^>]*>([^<]+)</p>', text)
   article = ' '.join(p.strip() for p in paras if len(p.strip()) > 30)
   

5. Format & Output — use the template below.

Output Template

📰 每日新闻速报 | YYYY年MM月DD日

@何旭

## 💰 财经要闻

**① 新闻标题**
正文摘要约300字，涵盖事件核心信息及各方反应。
*来源：新浪财经 | YYYY年MM月DD日*

## ⚔️ 军事动态

**① 新闻标题**
正文摘要约300字，涵盖事件核心信息及各方反应。
*来源：新浪军事 | YYYY年MM月DD日*

## 💻 科技前沿

**① 新闻标题**
正文摘要约300字，涵盖事件核心信息及各方反应。
*来源：新浪科技 | YYYY年MM月DD日*

Troubleshooting

Cron job 报 401 AuthenticationError / 群里只看到 401 或 01

这是模型/搜索服务 API Key 在 cron 子进程中不可见或失效，通常不是飞书凭证问题。最稳方案是绕开 LLM 调用：

• 对每日新闻简报，优先使用 no_agent=True 脚本模式，让脚本 stdout 直接作为推送内容，避免模型 API Key 401 导致任务失败。
• 当前可复用脚本路径：/root/.hermes/scripts/daily_news_briefing_noagent.py。它使用新浪 Feed API 拉取财经/科技，并从新闻流关键词筛选军事相关新闻；无需大模型或 Firecrawl Key。
• 更新 job 示例：cronjob(action='update', job_id='<id>', no_agent=True, script='daily_news_briefing_noagent.py', prompt='', skills=[], deliver='feishu:oc_YOUR_CHAT_ID')
• 对助手群投递不要长期用 origin 或 feishu:助手群，优先用明确 chat_id：feishu:oc_YOUR_CHAT_ID，避免 DM/群/topic 名称解析歧义。
• no_agent 脚本内部不要再直接调用飞书发送 API。脚本只 print() 最终正文，由 Hermes cron 统一投递；否则会出现脚本自发到错误 chat（如 DM）+ cron 再把 SUCCESS/ERROR 当正文发群的错乱。
• 不要尝试仅通过环境变量传递密钥 — cron 进程不稳定继承 .env 或 /etc/environment。
• 验证不能只看 last_status=ok。必须：
  1. 本地运行脚本确认 stdout 是用户要看的正文，不是 SUCCESS/ERROR/401；
  2. cronjob(action='run', job_id=...) 手动触发；
  3. 等 scheduler tick 后查 cronjob(action='list')，确认 last_status=ok 且 last_delivery_error=null；
  4. 查 ~/.hermes/logs/agent.log，确认 Job '<id>': delivered to feishu:oc_YOUR_CHAT_ID via live adapter。

See productivity/news-briefing skill for full cron 401 fix patterns (prompt embedding + no_agent script).

Cron job 报 "Response truncated due to output length limit"

这是模型生成的响应超出了输出长度限制，不是 API 或认证问题。症状：输出文件只有 ~1800-2000 字节，job 状态显示 error，last_error 为 RuntimeError: Response truncated due to output length limit。

修复方法（已验证有效）：

1. 严格限制新闻数量：每板块只选 2-3 条，不要选太多
2. 严格限制摘要长度：每条 150-200 字，不要写 300 字
3. 设置总字数上限：整个简报控制在 2500 字以内
4. 精简模板：移除冗余的说明文字和格式化符号
5. 附加技能：确保 cron job 关联了 daily-news-briefing 技能

Firecrawl returns old dates despite date in query

Firecrawl API /v2/search does not reliably filter by date. Searching "2026年6月13日 财经" may return articles from 2026-06-11 or earlier. Use Sina Feed API as primary for same-day news (lid=2516 finance, lid=2515 tech). Reserve Firecrawl for supplemental search or when Sina Feed API is unavailable.

Sina API returns empty items

• Try different lid values near the target range (e.g., 2514–2525)
• Try adding k={keyword} parameter for search-based retrieval
• If all lids return 0 items, fall back to lid=2516 (finance) and filter

Sina article URLs return "页面没有找到"

• This is a known recurring issue — many Sina article URLs are stale even when the article exists
• Workaround: use web_search(query="site:mil.news.sina.com.cn 2026年6月12日 军事") to find the live URL
• For military articles, URLs under /zonghe/ subdirectory are more reliable than those under other paths
• If web_extract fails, fall back to browser_navigate + browser_snapshot to read the article content directly

Military page renders as "页面没有找到"

• The rollxml endpoint is deprecated. Always use the HTML page (https://mil.news.sina.com.cn/) via browser.
• If the browser times out, retry — the page loads JS content and may need a moment.

Article text has encoding issues

• Sina uses GB2312 or GBK encoding on some pages. Use decode('gb2312', errors='ignore') or errors='ignore' with UTF-8.

Example Output

Finance 📈

① Kospi指数暴跌逾3% 地缘风险重创芯片股 韩国基准股指周三大幅下跌，芯片股延续跌势且受美国袭击伊朗等地缘政治因素影响……Kospi 200波动性指数周二首度突破90创下纪录。 来源：新浪财经 | 2026年6月10日

Military ⚔️

① 罗斯福号航母领衔参加2026环太军演 以罗斯福号航母为核心的海军编队已领衔参加2026年环太平洋军事演习……演习涵盖海上补给、反潜作战、多国舰艇协同等科目。 来源：新浪军事 | 2026年6月9日

Technology 💻

① iOS 27体验：App冷启动快30% 4K滑动加载提升70% WWDC 2026发布的iOS 27虽被吐槽"挤牙膏"，实测数据却显示其性能提升显著：App冷启动速度平均快30%…… 来源：苹果汇 / 新浪科技 | 2026年6月10日