paper-triggered-survey

Paper-Triggered Survey Skill

Overview

This skill performs a targeted literature survey anchored around a specific paper provided by the user. It combines:

• Paper extraction and analysis from multiple input formats (PDF, arXiv URL, tweet)
• Citation graph traversal to find citing and cited-by papers
• Search-based discovery of related work using extracted keywords
• Cross-domain exploration applying the paper's methodology to other domains
• Innovation proposal generation suggesting 2-3 extensions of the paper's work

The skill systematically maps the paper's research landscape, identifies gaps in cited work, and proposes novel research directions based on the paper's core contributions and methods.

Input Detection

The skill automatically detects the input type from $ARGUMENTS:

• PDF file — Argument ends in .pdf

  • Read using Read tool with pagination (pages 1-20, then 21-40 for long papers)
  • Extract title, authors, abstract, key sections, and references

• arXiv URL — Argument contains arxiv.org/abs/

  • Extract paper ID from URL (e.g., 2301.12345)
  • Use paper_search.py with --arxiv-id to fetch paper metadata and PDF

• Tweet URL — Argument contains twitter.com or x.com

  • Use WebFetch tool to extract tweet content
  • Parse for paper references (titles, arXiv IDs, DOIs)
  • Locate and load the referenced paper

• Search query — Any other format

  • Treat as paper title/topic search query
  • Run paper_search.py --query to find the paper
  • Load the first result or ask user to clarify

Setup

Before running the skill, install dependencies:

pip install -r "BASE_DIR/scripts/requirements.txt"

Required packages:

• arxiv — arXiv API access
• requests — HTTP requests for paper fetching
• bibtexparser — BibTeX parsing and generation
• semanticscholar — Semantic Scholar API integration

Workflow

Step 1: Detect Input Type and Extract Paper

1. Parse $ARGUMENTS to determine input type
2. For PDF: Read file with tool, extracting title, authors, abstract
3. For arXiv URL: Extract ID and use paper_search.py --arxiv-id <ID> to fetch
4. For tweet: Use WebFetch to extract content, then locate referenced paper
5. For search query: Use paper_search.py --query "<query>" --max-results 5 to find paper
6. Output: Confirmed paper metadata (title, authors, year, venue, arXiv ID, DOI)