enterprise-agent-builder

The enterprise-agent-builder is a specialized engineering skill designed for the 'Build' phase of the Listen-Decode-Build-Ship methodology. It acts as the primary bridge between conceptual agent specifications and fully deployable production software. This skill is intended for developers, system architects, and automation engineers who need to rapidly transition from a structured problem definition (provided by enterprise-problem-decoder) to a working codebase. By automating the architectural design process and boilerplate generation, it allows teams to focus on business logic and integration requirements rather than standard scaffold maintenance.

• Automatically generates a comprehensive Agent Blueprint, covering agent type, trigger mechanisms, system inputs, reasoning logic, and human-in-the-loop gates.

• Creates visual-style text-based data flow diagrams to map the movement of information between trigger sources, LLM reasoning cores, and target enterprise systems.

• Produces production-ready Python code integrated with the Anthropic Claude API, incorporating standardized tool definitions for fetching context, executing actions, and requesting human approval.

• Provides platform-specific adaptation logic, including support for Apex in Salesforce, JavaScript/Groovy for ServiceNow, and SQL-based querying for Snowflake environments.

• Enforces strict compliance and audit logging controls, ensuring that generated agents follow secure, enterprise-grade handling of PII/PHI and credential scopes.

• Input Requirements: The skill requires an agent_spec YAML file as its primary input. In the absence of a pre-existing spec, it will interactively prompt for critical parameters such as agent name, platform, industry compliance, and desired outcomes.

• Execution Strategy: The skill is built on the principle of 'no stubs'; the generated code is intended to be functional, deployable, and ready for integration testing upon the completion of the interaction.

• Architecture Focus: Emphasizes the separation of concerns between ingestion, reasoning, and action, ensuring that every agent built adheres to a clean, scalable architectural pattern.

• Compliance & Safety: Every blueprint includes explicit sections for audit logging and human-in-the-loop gates, mandating a 'human-approval-first' approach for all potentially irreversible actions like bulk deletions or financial updates.