The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from external sources (Amazon product reviews, descriptions, and metadata) via MCP services. This content is subsequently processed by the agent to generate research reports, creating a surface for indirect prompt injection attacks where malicious data could influence agent behavior.
Ingestion Points: Data is fetched from mcp.sorftime.com, mcp.sif.com, and mcp.xiyou.com using the product_reviews, product_detail, and keyword_detail methods as demonstrated in the Python implementation patterns in SKILL.md.
Boundary Markers: The instructions do not define delimiters or explicit 'ignore instructions' directives for the ingested data.
Capability Inventory: The skill performs file system writes to save research reports in local directories such as reports/ and category-reports/.
Sanitization: There is no evidence of input validation or sanitization for the retrieved Amazon content before it is incorporated into the agent's context.
[DATA_EXFILTRATION]: The skill transmits potentially sensitive business intelligence, such as specific product identifiers (ASINs) and target keywords, to third-party endpoints.
Evidence: Communication occurs with mcp.sorftime.com, mcp.sif.com, and mcp.xiyou.com via HTTP POST requests containing user-defined search parameters to facilitate research functions.

amazon-sorftime-research-mcp-skill