The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external data by fetching web page content using scripts/fetch_page.py. The instructions in SKILL.md and references/skill-contract.md provide clear guidance for the agent to treat scraped material as evidence only and to not follow any instructions embedded within that content. This mitigation reduces the risk of the agent being manipulated by instructions found on third-party websites.
[EXTERNAL_DOWNLOADS]: The scripts/generate_pdf_report.py script identifies an optional dependency on the reportlab Python library for PDF generation. The script does not automatically install this package but instead provides a clear error message and installation command for the user, ensuring the user remains in control of environment changes.
[COMMAND_EXECUTION]: The skill includes several Python scripts for automated SEO tasks. The scripts/fetch_page.py script implements strong security controls by validating requested URLs against a blocklist of private, loopback, and link-local IP addresses using socket.getaddrinfo. This effectively prevents SSRF attacks where the agent might be tricked into accessing internal services.

serp-markup-builder