The Agent Skills Directory

[SAFE]: The skill interacts exclusively with the vendor's official Model Context Protocol (MCP) server endpoint at https://api.wellapp.ai/v1/mcp for all data operations.- [SAFE]: A strict human-in-the-loop confirmation pattern is enforced, requiring the agent to display a full draft and receive explicit user approval before executing the well_create_invoice_from_data write tool.- [SAFE]: Robust instructional guardrails prevent the fabrication of financial data, ensuring that all amounts, tax IDs, and line items are sourced directly from user input or verified workspace defaults.- [SAFE]: The skill utilizes standard authentication and workspace resolution flows via the Well platform's native tools, adhering to principle of least privilege.

draft-invoice