AI Daily Digest skill
A /digest command skill: pulls articles from 90 tech blogs, scores them with an AI model and builds a daily Markdown digest with charts
- 1,659 ★ on the repository
- Last commit
Extended access
Why this level
- The script downloads RSS from 90 outside addresses and sends article titles and descriptions to Gemini or an OpenAI-style service
- The Gemini key travels in the request URL and is saved in plain text in config.json in the home folder
- The skill creates an output folder and a report file, and downloads the Bun runtime through npx
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/ai-daily-digestDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The script downloads RSS from 90 outside addresses and sends article titles and descriptions to Gemini or an OpenAI-style service
- The Gemini key travels in the request URL and is saved in plain text in config.json in the home folder
- The skill creates an output folder and a report file, and downloads the Bun runtime through npx
This is third-party code. Review the repository files before installing.
What it does
AI Daily Digest builds a daily digest from 90 technical blogs that Hacker News readers follow: simonwillison.net, paulgraham.com, overreacted.io, gwern.net, krebsonsecurity.com and others. The author takes the source list from the Hacker News Popularity Contest 2025 ranking and refers to a recommendation by Andrej Karpathy in the description. The skill starts with the `/digest` command, after which the agent asks for the period (24 hours, 48 hours, 72 hours or a week), the number of articles (10, 15 or 20), the output language and a Gemini key. The work is done by a single TypeScript file, `scripts/digest.ts`, with no third-party libraries. It downloads RSS and Atom from ten sources in parallel with a 15-second timeout, selects articles within the chosen period, sends them in batches of ten to the Gemini model to score on significance, quality and freshness and to sort into six categories, then writes a four to six sentence summary for the best ones with a translated title and a retelling of the day's trends. If Gemini fails, the script switches to any OpenAI-compatible service. The output is a Markdown file: the day's highlights, three must-read articles, a data table, Mermaid charts, a text chart and a tag cloud, then articles by category. Settings are saved in `~/.hn-daily-digest/config.json`.
Who it is for. For developers and engineering leads who follow English-language blogs and want a digest in Chinese or English in minutes.
Good fit when
- You want a morning list of the best technical articles of the last day or two
- You have no time to read dozens of blogs by hand and want headlines with summaries at once
- You need a Markdown report that opens in Obsidian or on GitHub along with Mermaid charts
- You want to see which topics (security, AI, engineering) dominate the blogs today
- You want to use DeepSeek or another model with an OpenAI-style interface instead of Gemini
Not a fit when
- You need your own sources: the 90 addresses are written in the script itself and the list cannot change without editing code
- You have no Gemini key and no key for a service with an OpenAI-style interface: the script does not work without one of them
- Articles must stay on your local machine: headlines and descriptions are sent to an outside model
- You need Russian output: the script knows only Chinese and English
- You need full article texts: the skill works with RSS descriptions
Example request
/digest for 48 hours, the 15 best articles, English output, save the report in outputLimitations
The author did not specify a license, rights to use and copy are not defined, check with the author.
The skill is written for OpenCode: its instructions contain `question()` calls that not every agent has. The script needs Bun (downloaded by `npx -y bun`) and network access.
The Gemini key goes in the URL of the request to the gemini-2.0-flash model, and when run through the skill it is saved in plain text in `config.json` in the home folder. By default the fallback uses gpt-4o-mini at api.openai.com; the address and model are changed with the `OPENAI_API_BASE` and `OPENAI_MODEL` variables.
At the start of every run the agent must print a promotional line about the "懂点儿AI" channel and its WeChat account. Sources that fail to open are skipped by the script.
The latest change in the repository dates from February 2026.
How to disable. Delete the skill folder from the agent's skills directory. The saved key and settings are in `~/.hn-daily-digest/config.json`; delete that folder yourself. Finished digests stay in the `output` folder or wherever `--output` pointed.
FAQ
Where do the 90 sources come from?
The author takes them from the Hacker News Popularity Contest 2025 ranking: independent blogs most popular with Hacker News readers. The full list is built into `scripts/digest.ts`.
How does the script score articles?
The model gives each article 1 to 10 on three criteria: significance, quality and freshness. It also assigns the article to one of six categories and picks keywords. Articles go to the model in batches of ten, two batches at a time.
What are the categories?
There are six: AI and machine learning, security, engineering, tools and open source, opinions and musings, and other.
What happens if Gemini fails or the quota runs out?
The script switches to an OpenAI-compatible interface. It needs `OPENAI_API_KEY`, optionally an `OPENAI_API_BASE` address (DeepSeek in the example) and a model in `OPENAI_MODEL`. If no model is set, one is chosen automatically.
How do I run the script without an agent?
Set `GEMINI_API_KEY`, then run `npx -y bun scripts/digest.ts --hours 48 --top-n 15 --lang zh --output ./digest.md`. The supported output languages are `zh` and `en`.
What if I want to replace Gemini with another provider?
The author gives a ready request for an agent: rewrite the address constant and the `callGemini` function in `scripts/digest.ts`, about 25 lines. The rest of the code does not depend on the provider. The scoring and summary prompts can stay unchanged.
Discussion
No questions or reviews yet.
Related
SkillsA library of 64 skills for founders: GTM, growth, SEO, email, graphics and video, installed with one npx command for Claude Code, Codex and more
AI Collab Playbook bundles 19 skills for Claude Code and Codex: paper review, Codex session recovery, anti-AI-tone editing and research-workflow helpers
Two skills for English prose: humanize rewrites a model draft in a natural voice, and ai-check flags machine-writing tells
SEO article writer skill with images
SEO Article Writer
A 10-stage Claude Code pipeline: Wordstat semantics, top-5 competitor analysis, AI cover and illustrations, publishing to WordPress, Dzen and Prisma Blog
Humanities Writing Companion walks a humanities paper through 12 modes: from sharpening the research question to a blind read and AI-use disclosure
Avoid AI Writing
avoid-ai-writing
A skill that finds typical machine-writing patterns in text and rewrites them, with detect-only and edit-in-place modes
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Repository stars | Main difference |
|---|---|---|---|
| AI Daily Digest (this entry) | no | 1,659 | |
| last30days | no | 65,797 | collects discussions of a topic over 30 days from Reddit, X, YouTube, HN, Polymarket and GitHub into a brief |
| Papr | no | 560 | an RSS reader with a CLI for agents |
README in short
The project description, in Chinese: a link to an article on how the skill was built, an overview picture, the dialog parameters and a command-line example. Then a five-step pipeline, the report structure, six categories and a list of strengths: a single file with no libraries, remembered configuration, charts. A separate section explains how to swap the model and shows addresses and variables for OpenAI, Anthropic, DeepSeek and Qwen. It ends with the list of sources naming popular bloggers.
README badge
Original SKILL.md
---
name: ai-daily-digest
description: "Fetches RSS feeds from 90 top Hacker News blogs (curated by Karpathy), uses AI to score and filter articles, and generates a daily digest in Markdown with Chinese-translated titles, category grouping, trend highlights, and visual statistics (Mermaid charts + tag cloud). Use when user mentions 'daily digest', 'RSS digest', 'blog digest', 'AI blogs', 'tech news summary', or asks to run /digest command. Trigger command: /digest."
---
# AI Daily Digest
从 Karpathy 推荐的 90 个热门技术博客中抓取最新文章，通过 AI 评分筛选，生成每日精选摘要。
## 命令
### `/digest`
运行每日摘要生成器。
**使用方式**: 输入 `/digest`，Agent 通过交互式引导收集参数后执行。
---
## 脚本目录
**重要**: 所有脚本位于此 skill 的 `scripts/` 子目录。
**Agent 执行说明**:
1. 确定此 SKILL.md 文件的目录路径为 `SKILL_DIR`
2. 脚本路径 = `${SKILL_DIR}/scripts/<script-name>.ts`
| 脚本 | 用途 |
|------|------|
| `scripts/digest.ts` | 主脚本 - RSS 抓取、AI 评分、生成摘要 |
---
## 配置持久化
配置文件路径: `~/.hn-daily-digest/config.json`
Agent 在执行前**必须检查**此文件是否存在：
1. 如果存在，读取并解析 JSON
2. 询问用户是否使用已保存配置
3. 执行完成后保存当前配置到此文件
**配置文件结构**:
```json
{