The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the eval "$(python ...)" pattern in SKILL.md to execute the output of Python scripts as shell commands. This includes a reference to ../ecommerce-listing/scripts/extract-listing-next-page.py, which is located outside the skill's own directory structure.
[DYNAMIC_EXECUTION]: The scripts/extract-reviews.py script generates JavaScript code at runtime by injecting configuration parameters into a string template. This generated code is subsequently executed within the browser context.
[PROMPT_INJECTION]: The skill is designed to ingest untrusted third-party content from e-commerce websites, creating a surface for indirect prompt injection.
Ingestion points: The scripts/extract-reviews.py script extracts text from various DOM elements (titles, bodies, and author names) across multiple platforms like Amazon and WooCommerce.
Boundary markers: The skill does not employ any boundary markers or instructions to isolate the scraped data from the agent's command context.
Capability inventory: The skill uses the bash tool to execute Python scripts which in turn generate and execute JavaScript in the browser.
Sanitization: There is no evidence of sanitization, filtering, or validation performed on the extracted review content before it is returned to the agent.

ecommerce-reviews