The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data from web pages, creating a surface for potential indirect prompt injection attacks.\n
Ingestion points: scripts/fetch_page.py retrieves content from arbitrary HTTP/HTTPS URLs for processing.\n
Boundary markers: Instructions in SKILL.md and references/skill-contract.md explicitly state that scraped content must be treated as untrusted and that embedded instructions should be ignored.\n
Capability inventory: The skill can perform network operations via scripts/fetch_page.py and write to the local file system using scripts/generate_markdown_report.py and scripts/schema_generator.py.\n
Sanitization: Instructions mandate treating external data as evidence only, preventing the agent from executing commands found within the data.\n- [EXTERNAL_DOWNLOADS]: The skill includes a script capable of fetching content from any public URL.\n
scripts/fetch_page.py: This utility allows for network fetching with built-in SSRF protections, such as resolving hostnames to IP addresses and blocking access to private, reserved, or loopback ranges.\n- [COMMAND_EXECUTION]: Several Python scripts are provided as CLI utilities for internal logic and report generation.\n
scripts/seo_health_score.py: Computes weighted SEO scores based on provided input parameters.\n
scripts/schema_generator.py: Fills JSON-LD templates with provided values, utilizing standard JSON escaping to prevent injection into the output artifacts.

seo-schema