The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches data from external sources, specifically completion.amazon.com for keyword suggestions and various Amazon regional domains (e.g., www.amazon.com, www.amazon.co.uk) to retrieve competitor product information. These downloads target well-known services directly related to the skill's primary purpose of Amazon PPC management.
[COMMAND_EXECUTION]: The skill includes a shell script (scripts/fetch-competitor.sh) that uses curl and text processing tools like grep and sed to extract product titles, brands, and categories from Amazon listing pages. This script is executed locally to provide the agent with necessary product context for campaign building.
[DYNAMIC_EXECUTION]: The instructions contain a command that pipes JSON data from an official Amazon API into a Python one-liner: python3 -c "import sys,json; [print(s['value']) for s in json.load(sys.stdin).get('suggestions',[])]". This is a standard data processing pattern where local Python code is used to parse structured data received from a well-known service.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content retrieved from the web (Amazon product listings) and user-supplied CSV search term reports. This creates an ingestion surface for potential indirect instructions. However, the skill treats this data as analytical input and lacks high-privilege capabilities or automated write actions that would enable a meaningful attack escalation. The risk is considered low and inherent to tools performing web-based research.

amazon-ppc-campaign