The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill generates JavaScript scripts at runtime using scripts/extract-search-results.py and executes the resulting code within a browser session via the browser-act eval command. This pattern of script generation and execution represents a dynamic code risk.
[INDIRECT_PROMPT_INJECTION]: The skill extracts product metadata, titles, and badge text directly from Amazon search result pages. Since this content is provided by external sellers, it represents an attack surface for indirect prompt injection if the agent subsequently processes the scraped text as instructions.
Ingestion points: SKILL.md and scripts/extract-search-results.py ingest data from Amazon product cards (specifically titles, badges, and delivery descriptions).
Boundary markers: No specific boundary markers or 'ignore' instructions are used when passing extracted data back to the agent.
Capability inventory: The agent has access to the browser-act tool, which allows for shell-like interaction with a browser environment, including navigation and code evaluation.
Sanitization: The extraction script performs basic numeric parsing but does not sanitize or escape the string content of product titles or badges before returning them in the JSON response.

amazon-search-listing