The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill generates structured funnel definitions by ingesting potentially untrusted data from merchant requests and store catalog metadata.
Ingestion points: Data is retrieved from lexsis_catalog, lexsis_pages, lexsis_discover, and user-supplied merchant requests (SKILL.md).
Boundary markers: No explicit delimiter or instruction-override protection is defined for data interpolation.
Capability inventory: The skill possesses write capabilities for storefront drafts (funnel_create, funnel_update) and the ability to open signed preview URLs in a browser (preview_funnel).
Sanitization: The skill incorporates a mandatory validate_funnel action to detect schema errors and unreachable logic before finalizing drafts, which acts as a technical control against malformed injections.
[EXTERNAL_DOWNLOADS]: The skill depends on a remote MCP server to interface with the Lexsis storefront platform.
Source: Fetches capabilities and executes actions via https://mcp.trylexsis.com/mcp (agents/openai.yaml).
Method: Utilizes the lexsis-ai tool with a streamable HTTP transport.

funnels