ArXiv Paper Search & Trend Analysis

Problem

When tracking new papers, engineers often need more than a list of arXiv links. They need a quick read on what a paper focuses on, where the novelty is, which directions appear hot in the current batch, and what trends are emerging. arxiv-paper-searcher treats search, abstract analysis, hotspot statistics, and trend judgment as one default workflow, rather than stopping at titles.

How It Works

The skill first parses keywords, result count, and sort order. If not specified, it defaults to 20 results, date sorting, and analysis based on titles and abstracts. Core outputs include:
- Original title, authors, publication time, focus hotspots, innovation assessment, and arXiv / PDF links for each paper
- Top 10 research directions in the current results
- 3 to 5 concrete trend observations

If the user confirms continuous tracking, the skill saves business settings such as keyword, max_results, sort, and timezone, then creates a scheduled task through OpenClaw cron and writes the jobId back to local config.json. Delivery is biased toward the current conversation, which suits updating results in the same chat.

Boundaries

Analysis must be based on abstracts, not title matching alone. Trend conclusions describe the current retrieved batch and should not be treated as a definitive view of the whole field. Scheduled monitoring is not the default step, and local config.json is not the scheduling source of truth. If the current conversation is deleted, delivery may fall back to a new conversation, interrupting context continuity.