The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from academic websites such as Google Scholar, JSTOR, and PubMed, as well as text extracted from academic papers.
Ingestion points: External content is fetched through the academic_stealth_fetch.sh script, the read_paper tool, and Exa search results.
Capability inventory: The skill utilizes shell script execution (academic_stealth_fetch.sh), network operations (Exa search and Obscura fetching), and file system writes for paper storage (~/.arxiv-papers).
Boundary markers: There are no explicit delimiters or instructions within the skill to isolate or flag potential instructions embedded in the scraped data.
Sanitization: Data is extracted using JavaScript selectors but is returned to the agent without specific sanitization to filter out malicious prompt content.
[COMMAND_EXECUTION]: The skill executes a custom shell script, academic_stealth_fetch.sh, which triggers the obscura CLI tool to scrape bot-protected websites.
[EXTERNAL_DOWNLOADS]: The documentation provides instructions for installing third-party tools, including arxiv-mcp-server, via package managers like uv and npx.
[DYNAMIC_EXECUTION]: The web scraping logic employs the --eval functionality of the obscura browser to execute JavaScript strings for targeted data extraction from the DOM at runtime.

academic-research