Package ai.tabforge.workshop.agents

package ai.tabforge.workshop.agents
Concrete sub-agent implementations for the Workshop review pipeline.

Each class in this package is a specialist that calls the Anthropic Java SDK with a domain-specific system prompt to analyze a specific aspect of Java code:

  • SecurityAuditorAgent — finds security vulnerabilities in JAX-RS / JPA
  • TransactionAnalystAgent — finds Jakarta EE transaction boundary errors
  • PerformanceAnalystAgent — finds JPA N+1 queries and EJB lifecycle issues
  • ArchitectureCheckerAgent — finds layer violations and CDI dependency issues
  • SelfEvaluatorAgent — meta-agent that evaluates other agents' output

CERTIFICATION NOTE — Agentic Architecture & Orchestration (27% of exam): Narrow specialist agents with focused system prompts outperform a single general-purpose agent on complex multi-domain analysis. Agent specialization is a core pattern in production agentic systems.

Implemented in Phase 1 Day 5-6 (Security) and Phase 2 Day 11-14 (all agents).