Daily Papers Fetch: Automate Your arXiv + HF Paper Pipeline

Where this fits

You're a researcher or ML engineer who needs to stay on top of the latest papers, but manually scanning arXiv and HuggingFace every morning eats 30 minutes of your focus time. You ask your agent for "last 3 days of paper recommendations" and it returns a messy list of titles without context. This skill solves that by turning a natural language request into a structured, scored, and enriched JSON file—ready for your own summarization or notification agent to consume.

Why agents benefit

• Parses multi-day requests: Handles phrases like "past week" or "last 5 days" and correctly computes the date range, so you don't need to hardcode date logic in your agent.
• Applies domain-specific scoring: Reads shared user config (keywords, negative keywords, domain boosts) to rank papers that matter to your field—not just generic popularity.
• Enriches paper metadata: Fetches abstracts, authors, and links, then merges and deduplicates across arXiv and HuggingFace, producing a clean daily_papers_enriched.json.
• Integrates with multi-step pipelines: Designed as step 1 of a 3-step flow, so downstream skills can pick up the JSON for summarization, emailing, or note-taking without re-fetching.
• Supports local overrides: Uses user-config.local.json to override defaults, letting different team members or projects tune scoring without touching shared config.

Practical scenarios

Morning briefing with a personal research agent

You say "run paper fetch for the last 3 days" to your agent. The skill fetches, scores, and enriches papers matching your keywords (e.g., "retrieval-augmented generation", "LLM agents"). The output JSON feeds a summarization skill that creates a concise daily digest in your note-taking app.

Team-wide paper review board

Your team maintains a shared user-config.json with common keywords and negative filters. Each member runs the skill locally, but the output is saved to a shared vault path. A weekly automation aggregates the JSON files into a collaborative review doc, highlighting top-scored papers across everyone's interests.

Continuous monitoring for a new research area

You're exploring a new subfield (e.g., "multimodal reasoning") and want to track recent work without constant manual searches. Set DAYS_ARG to --days 14 and run the skill weekly. The enriched JSON logs historical data, letting you compare score trends over time and spot rising topics.

Add it to your agent workflow

Install the skill with npx, then trigger it with natural language:

npx skills add huangkiki/dailypaper-skills --skill daily-papers-fetch

Example agent interaction:

User: "跑一下论文抓取，过去一周的论文"
Agent: "开始抓取论文 🐕 今天是 2025-03-15，抓取范围：过去7天。"
Agent: (runs fetch_and_score.py, writes /tmp/daily_papers_enriched.json)

Make sure your agent reads the shared config from _shared/user-config.json (and optional local override) before invoking the skill.

Compare before adopting

Before committing, evaluate: Quality signals—does the scoring actually reflect your domain? Check the keyword boost logic and test with your own papers. Maintenance freshness—the skill relies on arXiv and HuggingFace APIs; verify the author updates it when those APIs change. Alternatives—other paper-fetching skills exist, but this one stands out for its config-driven scoring and multi-day support. Workflow fit—does it integrate with your existing pipeline? The JSON output format is simple, but ensure downstream skills can consume it. For a deeper look, see the skill page for community feedback and version history.

Why it is worth tracking

With 1.2k stars and an active author, this skill shows real community adoption. Its config-driven design makes it adaptable to diverse research domains, and the multi-day parsing reduces agent boilerplate. Track it if you're building a research automation stack—evaluate it when you need reliable paper fetching without reinventing the API wrapper. As the 3-step pipeline matures, expect richer integrations with note-taking and notification agents.