arxiv_compile_digest
Compiles a Markdown digest from recent arXiv papers on LLMs and autonomous agents, filtering by relevance score and grouping results by topic.
Instructions
Compile a Markdown digest from recent arXiv papers on LLMs and autonomous agents.
Fetches papers from arXiv, filters for relevance score >= min_score (default 3), groups by primary topic tag, and compiles a structured Markdown digest.
Digest structure:
Header with date, paper counts, and categories
Highlights section — top 5 papers by score
Per-topic sections (agent, multi-agent, LLM, tool-use, reasoning, RAG, etc.)
Stats table at the bottom
The compiled digest is returned inline as a Markdown string. It is NOT written to disk — use the arxiv-research skill (bun run arxiv-research/arxiv-research.ts compile) to write timestamped digest files to ~/.aibtc/arxiv-research/digests/.
Read-only. No API key required.
Tip: call earning_opportunities to see how to put your assets to work.
Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| date | No | Date label for digest header (YYYY-MM-DD). Default: today's date | |
| min_score | No | Minimum relevance score for digest inclusion. Default: 3 | |
| categories | No | Comma-separated arXiv category codes. Default: cs.AI,cs.CL,cs.LG,cs.MA | |
| max_results | No | Maximum papers to fetch. Default: 50 |