The Agent Skills Directory

[SAFE]: The skill defines a robust workflow for interacting with the amazon_ads MCP server to generate and analyze Sponsored Products reports. It contains no self-executing code, external downloads, or persistence mechanisms.
[INDIRECT_PROMPT_INJECTION]: The skill has a data processing surface as it ingests CSV report content via the read_download tool (noted in references/mcp-workflow.md). It maintains security through defined ingestion points and clear boundary markers by validating expected v1 dotted headers (e.g., searchTerm.value). Capabilities are confined to standard advertising API tools (CreateReport, RetrieveReport, QueryAdvertiserAccount). The skill implements sanitization through explicit type casting (float/int) and regex partitioning for ASIN detection, which prevents data confusion during processing.
[COMMAND_EXECUTION]: All shell-like tool invocations are mapped to the intended advertising management functionality. No attempts at privilege escalation, unauthorized file system access, or data exfiltration were found.

amazon-ads-sp-search-term-explorer