zero-finance

This skill provides a complete AI-driven email pipeline designed for the 0 Finance ecosystem, enabling autonomous management of financial communications. It functions as a conversational agent that treats every email interaction as part of an ongoing session rather than a rigid state machine, allowing the AI to understand complex user intent, clarifications, and follow-ups. By leveraging advanced language models, the system handles parsing, tool execution, and response generation without relying on brittle regex patterns or pre-defined branching logic.

The system is engineered to integrate directly with email providers like Resend and SES, mapping inbound webhooks to specific session states. It performs critical financial operations including generating and sending invoice drafts, proposing ACH/IBAN bank transfers, and tracking account balances. Attachments such as PDF invoices and receipt images are handled natively as file parts, ensuring that the model maintains full context across the entire conversation history, which is persisted for long-term tracking and auditability.

• Automated extraction and reconciliation of invoice details directly from raw email content and attachments.

• End-to-end management of financial workflows including create, update, and send actions for invoices.

• Real-time bank interaction capabilities, such as checking balances, listing saved accounts, and proposing transfers with approval gates.

• Context-aware conversation handling that processes corrections like "no, change the amount to $600" or "send to this recipient instead" gracefully.

• Secure integration with Resend and AWS SES, utilizing environment-based credentials for workspace-specific operations.

• Inputs include inbound webhook payloads, existing session states, and workspace mapping configurations.

• Outputs consist of outbound email replies, updated session states, and triggered tool calls for financial services.

• Designed for high-reliability agent operations; if provider credentials are missing, the agent will pause and request setup.

• The system adheres to the 'CLI-first banking' philosophy, ensuring that AI-generated financial data remains consistent with user terminal commands and dashboard status.

• The architecture is non-branching: all logic is delegated to the LLM to ensure maximum flexibility in handling human nuances and conversational corrections.