The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local Python and Bash scripts located in the scripts/ directory to parse Amazon accessibility snapshots and rank products based on user criteria.
[EXTERNAL_DOWNLOADS]: Uses the agent-browser CLI to access amazon.com, a well-known and trusted service, to retrieve product information.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes untrusted data from Amazon search results. Maliciously crafted product information could attempt to bias the agent's recommendations.
Ingestion points: Amazon search result snapshots and individual product page data.
Boundary markers: Absent. The skill does not use specific delimiters or instructions to ignore potential commands within the scraped web content.
Capability inventory: Access to agent-browser for web navigation and the ability to execute local scripts via bash and python3.
Sanitization: The skill uses regex and jq for structured data extraction, which mitigates risk by treating most external content as data fields rather than direct instructions.

amazon-shopping