The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill retrieves academic metadata and paper information from the official APIs of arXiv (export.arxiv.org) and Semantic Scholar (api.semanticscholar.org). These are well-known and trusted academic research services.\n- [COMMAND_EXECUTION]: Provides shell command examples using curl and python3. While automated scans flagged these as potential remote code execution patterns due to shell pipes, manual analysis confirms they are used for parsing XML/JSON data with inline Python scripts or local files. The skill treats the remote response as data input for processing rather than executing it as a script.\n- [PROMPT_INJECTION]: The skill processes untrusted text (abstracts and titles) from external repositories, which presents a surface for indirect prompt injection. This is a low-risk finding typical of research tools that ingestion external text content; risk is mitigated by the use of structured XML/JSON parsing.\n- [SAFE]: The provided helper script scripts/search_arxiv.py and the Markdown code snippets use only Python standard libraries for data processing and correctly sanitize search parameters before API requests.

arxiv