ai-daily-digest
3
总安装量
2
周安装量
#58883
全站排名
安装命令
npx skills add https://github.com/harrishan/ai-daily-digest --skill ai-daily-digest
Agent 安装分布
pi
2
gemini-cli
2
github-copilot
2
codex
2
kimi-cli
2
cursor
2
Skill 文档
Tech Daily Digest
Generate a structured daily digest from 90+ top tech blogs curated from Hacker News Popularity Contest 2025.
Trigger
User says /digest or asks for a tech news digest/daily briefing.
Parameters (ask user if not specified)
| Param | Options | Default |
|---|---|---|
| Time range | 24h / 48h / 72h | 24h |
| Top N articles | 10 / 15 / 20 | 15 |
| Language | zh / en | zh |
Workflow
Step 1: Fetch RSS feeds
Run the bundled fetcher script:
node <skill_dir>/scripts/fetch-rss.mjs --hours <HOURS> --sources <skill_dir>/references/sources.json
This outputs JSON to stdout. Capture it. The script:
- Fetches 92 RSS/Atom feeds concurrently (15 parallel, 15s timeout)
- Parses XML (RSS 2.0 + Atom)
- Filters articles by time window
- Outputs sorted JSON array of
{title, link, summary, date, source, sourceUrl}
Step 2: Score and classify
From the fetched articles JSON, score each article on three dimensions (1-10):
- Relevance â How relevant to AI, software engineering, security, or tech industry
- Quality â Depth of insight, originality, technical substance
- Timeliness â Breaking news or emerging trend vs. evergreen content
Classify into categories:
- ð¤ AI / ML
- ð Security
- âï¸ Engineering
- ð Tools / Open Source
- ð¡ Opinion / Essay
- ð Other
Select the top N articles by total score.
Step 3: Generate summaries
For each selected article:
- If
summaryfrom RSS is sufficient (>100 chars), use it as basis - If not, use
web_fetchto read the full article - Generate a structured summary:
- Chinese title translation (keep original as link text)
- 2-3 sentence summary covering: core problem â key insight â conclusion
- Recommendation reason (1 sentence)
- Keywords (2-3 tags)
Step 4: Generate trend highlights
Analyze all selected articles together and identify 2-3 macro trends in the tech world today.
Step 5: Format output
Output as a Telegram-friendly message (not a file):
ð° Tech Daily Digest â {date}
æ¥èª Karpathy æ¨èç 92 ä¸ªé¡¶çº§ææ¯å客
ð 仿¥çç¹
{2-3 sentence macro trend summary}
ð 仿¥å¿
读 (Top 3)
1. {Chinese title}
{source} · {relative time}
{summary}
ð·ï¸ {keywords}
2. ...
3. ...
ð æ´å¤ç²¾é
4. {Chinese title} â {source} · {one-line summary}
5. ...
...
ð ç»è®¡ï¼{N} æº â {M} ç¯ â {K} ç¯ç²¾é
Notes
- No external API key needed. The agent itself handles scoring and summarization.
- RSS sources are from Karpathy’s curated list of Hacker News top blogs.
- If
execis unavailable, fall back to usingweb_fetchon individual RSS URLs (slower but works). - The script requires Node.js 18+ (available on all OpenClaw instances).
- Source list can be customized by editing
references/sources.json.