arxiv-digest

Arxiv Digest (L3 - Autonomous)

You are executing the arxiv-digest skill.

User Request

$ARGUMENTS

Purpose

Automatically fetch, filter, and summarize arxiv papers matching your research interests.

Autonomy Level: L3 (Autonomous)

• Human: Set interests once, review final digest
• AI: Search → filter → summarize → validate → present

Agent Loop

while not goal_reached:
    main_agent.plan(current_state)
    researcher.fetch_arxiv(categories, keywords)
    prover.filter_by_relevance(papers, interests)
    writer.summarize(relevant_papers)
    polisher.format_digest()
    if digest_ready:
        present_to_human(digest)

Input Required

• arxiv categories (e.g., quant-ph, cs.LG, cond-mat)
• Keywords/topics of interest
• Digest frequency (daily/weekly)

Output Format

# Arxiv Digest - [Date]

## High Relevance
### [Paper Title](arxiv_link)
**Authors**: ...
**Abstract**: 2-3 sentence summary
**Why relevant**: connection to your interests

## Medium Relevance
...

## Quick Mentions
- [Title](link) - one-line summary

Human Checkpoint

• Review digest, mark papers to read in full
• Optionally refine interest keywords

Categories

• quant-ph - Quantum Physics
• cs.LG - Machine Learning
• cond-mat - Condensed Matter
• hep-th - High Energy Physics Theory
• stat.ML - Statistics ML

Tool

arxiv_fetch.py - CLI to fetch papers from arxiv API.

# Fetch ML papers
./arxiv_fetch.py -c cs.LG -n 10 -f markdown

# Search by keywords
./arxiv_fetch.py -k "transformer" "attention" -f json