Package ai.tabforge.workshop.model


package ai.tabforge.workshop.model
Domain model for the Claude Architect Workshop MCP server.

These are pure Java records — no framework dependencies, no SDK imports. They represent the data structures that flow between all layers of the system:

  • Finding — a single issue discovered by a sub-agent
  • ReviewReport — the complete aggregated output of a review
  • ReviewScope — what to review (changed files vs full project)
  • AgentResult — what a sub-agent returns after calling Claude API
  • EscalationRequest — sent to Claude Desktop when human input is needed
  • HumanDecision — the developer's response to an escalation
  • Severity — CRITICAL / WARNING / INFO

CERTIFICATION NOTE — Prompt Engineering & Structured Output (20% of exam): The JSON schema enforced in every Anthropic API call is derived from these record types. Finding IS the output contract that every sub-agent's response_format parameter enforces.

Implemented in Phase 1 Day 3-4.