Add arxiv research skill

[Ashish-Soni08]

ArXiv paper research tools for Zo Computer.

Implements the same functionality as arxiv-mcp-server — but as a native Zo skill instead of an MCP server. Both tools use arXiv's public API (export.arxiv.org).

Why a native skill?

• MCP servers require running a separate Python process with JSON-RPC communication overhead
• Zo skills run natively with Bun/TypeScript — no extra processes, no protocol translation
• Same storage path (~/.arxiv-mcp-server/papers/), same results, zero additional setup

Features

• search_papers: Search arXiv with date/category filters
• download_paper: Download papers by ID
• convert_paper: PDF → Markdown with page markers for citations
• list_papers: List downloaded papers with [MD] indicator
• read_paper: Auto-detects format (markdown if available, else text)
• deep-paper-analysis: One-command analysis workflow

Usage

bun arxiv.ts search "transformer architecture" --max 5
bun arxiv.ts download 2401.12345
bun arxiv.ts convert 2401.12345  # Creates markdown
bun arxiv.ts read 2401.12345     # Reads markdown if converted

[skeletor-js]

This is fairly close. I would suggest a cleanup pass before merge.

• Reconsider the hidden local storage default under ~/.arxiv-mcp-server.
• Clean up the inline Python fallback flow for PDF extraction and conversion.

After that, this looks much closer to mergeable.

[Ashish-Soni08]

@skeletor-js Thanks for the feedback! I've gone ahead and made the suggested cleanup passes:

Storage Path: Changed the local storage default from ~/.arxiv-mcp-server/papers/ to a more appropriate /home/workspace/Research/arxiv-papers.
Inline Python Removed: Extracted the inline Python fallback flows into clean, standalone scripts (extract_text.py and convert_to_markdown.py).
Let me know if this looks good to merge now or if there is any other feedback!