arxiv-research-secure（arxiv 研究安全）

ArXiv Research Assistant Secure

name: arxiv-research-secure description: Advanced ArXiv paper search with local caching, smart summarization, and research tracking. Secure alternative with no shell execution.

ArXiv Research Assistant Secure

Overview

Advanced research tool for ArXiv papers:

• Smart search — Multi-keyword, author, category filtering
• Local caching — Avoid repeated API calls, offline reading
• AI summarization — Automatic abstract extraction with LLM
• Research tracking — Log papers to structured markdown
• PDF download — Local storage with metadata
• Zero shell — Pure Python, no curl/wget/exec

Security Model

Network Isolation

# Only HTTPS to export.arxiv.org
ALLOWED_HOSTS = ["export.arxiv.org", "arxiv.org"]
TIMEOUT_SECONDS = 30
MAX_RETRIES = 3

Input Validation

def validate_query(query: str) -> str:
    # Block injection attempts
    FORBIDDEN = [";", "|", "&", "$", "`", "\"", "'", "<", ">", "..", "//"]
    for char in FORBIDDEN:
        if char in query:
            raise ValueError(f"Invalid character in query: {char}")
    return query[:200]  # Max 200 chars

Local Storage Only

CACHE_DIR = WORKSPACE / ".arxiv_cache"
PAPERS_DIR = CACHE_DIR / "papers"
METADATA_DIR = CACHE_DIR / "metadata"

# No external network after download
# All operations on local files

Capabilities

1. Search Papers

arxiv-secure search "transformer energy consumption" --max=10 --sort=relevance
arxiv-secure search "author:LeCun" --category=cs.LG
arxiv-secure search "LLM reasoning" --date-from=2024-01-01

2. Download & Cache

arxiv-secure fetch 2501.12345           # Download by ID
arxiv-secure fetch --search="query" --auto-download  # Download all results

3. Smart Summary

arxiv-secure summarize 2501.12345       # AI summary of paper
arxiv-secure summarize --file=paper.pdf # Summarize local PDF

4. Research Log

arxiv-secure log 2501.12345             # Add to research log
arxiv-secure log --search="query" --auto-log
arxiv-secure list-log                   # Show research history
arxiv-secure export-log --format=md     # Export for reports

5. Manage Cache

arxiv-secure cache-list                 # List cached papers
arxiv-secure cache-clear --older-than=30d
arxiv-secure cache-stats                # Disk usage

6. Batch Operations

arxiv-secure batch --search="volatility modeling" --max=20 --download --summarize --log

Workflow Examples

Daily Research Digest

# Morning: Check new papers on your topics
arxiv-secure search "quantitative finance volatility" --date-from=yesterday --max=5 --summarize

# Log interesting ones
arxiv-secure log 2501.12345
arxiv-secure log 2501.12346

# Export weekly report
arxiv-secure export-log --format=md --since=last-week > weekly_report.md

Deep Research Session

# Search broadly
arxiv-secure search "transformer energy efficiency" --max=50

# Download all
arxiv-secure batch --search="transformer energy efficiency" --max=20 --download

# Summarize batch
arxiv-secure batch --ids=2501.12345,2501.12346,2501.12347 --summarize

# Log to research tracker
arxiv-secure batch --ids=2501.12345,2501.12346 --log

Output Formats

Console Table

ID          Title                          Authors        Date     Category
2501.12345  Energy-Efficient LLMs          Smith et al.   2025-01  cs.LG
2501.12346  Transformer Optimization         Chen et al.    2025-01  cs.CL

Markdown Log

## [2025-01-15] Research: Energy-Efficient LLMs
- **Paper**: Energy-Efficient Transformers for Edge Devices
- **Authors**: Smith, J., Chen, L., Kumar, R.
- **arXiv**: 2501.12345
- **Category**: cs.LG (Computation and Language)
- **Summary**: Proposes a novel pruning technique...
- **Relevance**: High - directly applicable to RTE volatility prediction models
- **Downloaded**: ✅ paper_2501.12345.pdf

JSON Export

{
  "query": "transformer energy",
  "date": "2025-01-15",
  "papers": [
    {"id": "2501.12345", "title": "...", "summary": "...", "relevance": 0.85}
  ]
}

Resources

scripts/

• arxiv_client.py — Secure API client with caching
• paper_summarizer.py — LLM-based summarization
• research_logger.py — Structured logging to markdown
• pdf_downloader.py — Safe PDF download and storage

references/

• arxiv_api_reference.md — API documentation
• research_templates.md — Log templates and formats

assets/

• paper_template.md — Default paper log template