Introduction

To answer the question “What happened in the AI world today?”, agents typically resort to building their own crawlers or integrating multiple authenticated news APIs. While AITOP (aitop.news) already aggregates, summarizes, and curates daily content from multiple sources, integrating this public data into an Agent still involves handling endpoints, pagination, and link display standards. Below is an introduction to AI News AI TOP NEWS on SkillHub (Coordinate: @user_507a5056/airebao, current version 1.0.8): a skill designed to read AITOP public content. It does not require an API Key (tagged requires_api_key=false) and does not depend on additional MCP configurations.

What is this

AI News AI TOP NEWS targets requests such as “querying, searching, and expanding AITOP’s AI industry dynamics and daily curated reports.” It covers today’s hot topics, recent models/products/papers/industry dynamics, keyword searches, AITOP daily reports for specific dates, news details, as well as source verification and original text citations.

The data comes from a community-published Skill (Source: community), categorized as knowledge-management, and published by user_507a5056. The official installation package includes aitop/SKILL.md (routing, schema, output templates, and prohibited behaviors) and aitop/scripts/fetch_aitop.py (a fetch script using only the Python standard library).

In one sentence: It allows Agents to access AITOP public interfaces without an API Key or MCP configuration, and output verifiable results according to a “prioritize in-site details, verify original text on demand” linking standard.

Core Features

Four Anonymous Public Endpoints

The skill covers the following GET public paths of AITOP (all are anonymous access in the documentation, no token required):

Path Function
/api/public/items List of AI dynamics; supports mode (selected / all), since, category, q filters, and cursor pagination
/api/public/daily, /api/public/daily/{YYYY-MM-DD} Latest or specified date’s curated daily report
/api/public/dailies List of dates with existing daily reports (discovery)
/api/public/site-config Real-time site configuration, used to verify categories, regions, and other enums

Smart Routing

Broad questions follow a “rolling time window” to fetch dynamics, rather than defaulting to a daily report. The /daily endpoint is used only when the user explicitly says “daily report”; mode=all is used only when “all” is specified. This prevents mixing daily reports (UTC 0:00 slices) with rolling time windows.

  • detailUrl: AITOP in-site detail page (/item/{id}), serving as the default display link for users.
  • sourceUrl: Upstream original text, provided only for citations, fact-checking, or when the user asks to “see the original”.
  • url: Compatibility alias for sourceUrl, marked as deprecated; new logic should use sourceUrl exclusively.

When summaries are generated by the model, the skill requires a prompt: “Please verify important information against the sourceUrl”; do not fabricate fields not returned by the API, and do not treat summaries as citations of the original text.

Engineering Conventions (Scripts & Pagination)

The built-in fetch_aitop.py handles the 403 response returned by public interfaces for the default curl User-Agent: the script embeds a real UA and provides readable diagnostics for 403 errors.

Pagination must be serial: stop when hasMore=false; concurrent pagination is prohibited. Instead of “fetching everything and filtering client-side,” use server-side filtering (?q=, ?since=, ?category=).

Enums such as categories should be based on the real-time return of /api/public/site-config to avoid hardcoding outdated values.

Boundaries and Extensions

The public documentation lists endpoints requiring Bearer tokens (such as /briefing, /compare, /tracking, /event/{id}, etc.); the skill does not store credentials for users and currently focuses on public endpoints. The roadmap includes integrating capabilities requiring temporary tokens, RSS/scheduled push, and rendering grouped by channel.

Installation and Activation

Open the skill page on SkillHub and install it (the documentation does not provide client-specific CLI commands; use the official coordinates below):

Once enabled, no API Key configuration is required. The package accesses the public API via aitop/scripts/fetch_aitop.py and does not require installing third-party Python packages.

Typical Usage

Triggered by natural language from the user; typical phrasing is defined in the skill’s frontmatter, for example:

  1. “Any new moves in the AI world today?”
  2. “What are the AI models/products/papers in the last 3 days?”
  3. “Give me an AI daily report” or “What did the AITOP daily report on May 13 cover?” (Date format follows user intent; the API path is /api/public/daily/{YYYY-MM-DD})
  4. “What has OpenAI released recently?” (Keyword search, using filters like q in items)
  5. “What’s new in AI?” / “AI hotspots or trends”

The Agent selects the endpoint according to routing decisions in SKILL.md, uses the fetch script to retrieve data, and outputs results with detailUrl / sourceUrl according to the template.

When fetching for local debugging, using curl directly with an improper UA may result in a 403; you should use the script included in the package or follow the UA instructions in the skill.

Applicable Scenarios and Notes

Who it is for

  • Content operators and公众号 (WeChat Official Account) authors: Use AITOP aggregated results for topic selection and source verification.
  • Researchers and product managers: Scan models, products, papers, and industry dynamics by category and time window.
  • Agent users who want to “ask in one sentence what happened in AI today”: Zero configuration, instant answers.

Things to Note

  • The semantics of daily reports and “dynamics from the last N days” are different; rely on the routing rules within the skill and do not mix them.
  • Factual statements should be verified against the sourceUrl; do not draw conclusions based solely on summaries.
  • Authenticated advanced capabilities are not within the scope of the current public skill; do not assume they are integrated.

Conclusion

AI News AI TOP NEWS consolidates link semantics, non-fabrication constraints, and pagination discipline into a single Skill, rather than scattered crawler scripts: ready to use upon installation, anonymous public endpoints, and standard library fetch. If you want to read AITOP stably within an Agent, you can obtain and install it from the official address:

https://api.skillhub.cn/user_507a5056/airebao