Agent-Agnostic Guardrails: Universal Java Code Quality with AGENTS.MD and Static Analysis
summary
Live demo builds a pipeline that feeds static-analysis results back into agents so generated Java code meets enterprise standards: agent output initially fails checks, then adding AGENTS.md plus integrated tools produces an ~85% drop in violations and a 60% rise in unit test coverage. The setup uses IDE agent plugins (Copilot/IDEA/AI Assistant and a Claude workaround), CI step mvn verify to run Checkstyle, SpotBugs, ArchUnit, and layered agents.md files; the loop materially improves quality.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 1:08Agent consistency and static analysis
- 5:25Solution: agents.md and context engineering
- 11:02Context windows and control
- 15:17Structure of agents.md files
- 18:07Static analysis tools
- 25:13Demo without guardrails
- 29:56Demo with agents.md and static analysis
- 34:30Switching coding agents
- 37:26Key takeaways
- 40:19Human oversight
- 41:50Q&A
related talks
11mAutomated Code Review for AI Agents
50mPanel: Beyond The Vibe: Code Quality First
30mArchitecting Integrity: Responsible Software Evolution in the AI Era
25mFrom AI Chaos to Production Ready: Using AI Agents to Improve Your Code Quality
59mTools & Tips to Modernize Your Java code with AI
30mAgents Building Agents
40mHow To Tame Your Agents?
30mOpen Challenges for Programming Agents