| name | arxiv-search-fix |
| description | Workaround for arxiv-search SSL issues when using python httpx |
| category | ai_collection |
arxiv-search-fix
Problem
The arxiv-search skill's internal httpx method fails with SSL errors:
[SSL: UNEXPECTED_EOF_WHILE_READING] EOF occurred in violation of protocol (_ssl.c:1032)
This happens both with and without the HTTP proxy.
Solution
Use curl with the HTTP proxy explicitly as shown below. This bypasses the problematic httpx usage.
curl -x http://127.0.0.1:7890 -G "https://export.arxiv.org/api/query" \
--data-urlencode "search_query=all:\"neuroscience\" AND submittedDate:[202607150000 TO 202607162359]" \
--data-urlencode "start=0" \
--data-urlencode "max_results=50"
Replace the search query and date range as needed.
Usage
When the arxiv-search skill fails due to SSL errors, fall back to the above curl command to fetch arXiv API results directly.
Extended Guidance for Automated Workflows
For scheduled cron jobs and automated research pipelines, consider these additional strategies:
Rate Limiting Mitigation
- Category-based queries: Use
cat:q-bio.NC OR cat:cs.NE instead of keyword searches to reduce request complexity
- Exponential backoff: Implement delays between retry attempts (start with 30s, double each retry)
- Local fallback: Check existing local repository INDEX.md files when live fetching fails
- Publishing cycles: Account for academic publishing patterns (fewer papers on weekends/holidays)
Security and Environment Compliance
- Always use HTTPS: Plain HTTP requests may be blocked by security scanners
- Explicit proxy configuration: Include
-x http://proxy:port or set http_proxy/https_proxy env vars
- Cron job constraints: In scheduled contexts,
execute_code is typically disabled; rely on terminal-based approaches
Monitoring Strategy
- Multi-source verification: Cross-check results across different query methods
- Graceful degradation: Design workflows that can proceed with partial data when full retrieval fails
- Error pattern recognition: Distinguish between transient errors (retryable) and systematic failures (require workflow changes)
