My personal notes on using AI coding assistants as a Java developer.
- How the runtime actually works — agent loop, tools, context limits — so failures feel debuggable instead of mystical.
- Project-level wiring —
AGENTS.md, SKILL.md, slash commands, subagents — so the assistant shows up already aligned with my architecture and standards.
- Java / Spring / DevOps reality — migrations, CI, Helm, K8s.
| Layer |
Technology |
| Language |
Java 21+ (records, sealed classes, virtual threads) |
| Framework |
Spring Boot 3.5+, Spring Data JDBC |
| Build |
Maven 3.9+ |
| CI/CD |
GitLab CI |
| Orchestration |
Kubernetes + Helm |
| Databases |
PostgreSQL, Oracle |
| Assistants |
Cursor (+ CLI), OpenCode |
Ready-to-use files in templates/:
| File |
Description |
AGENTS.md |
Project context template |
SKILL.md |
Skill template |
context-map.md |
Context map template |
team-onboarding.md |
Team onboarding guide |
command-cursor.md |
Cursor slash command template |
command-opencode.md |
OpenCode command template |
mcp-config-cursor.json |
MCP config for Cursor |
mcp-config-opencode.json |
MCP config for OpenCode |
liquibase-changelog.xml |
Liquibase changelog template |
sample-project/ |
Reference Maven project with skills, commands, and AGENTS.md — demo target for OpenCode workflows |
context-map.md and team-onboarding.md are canonical templates; sample-project/ holds project-specific overrides (paths, skill lists) where needed.
opencode-agent-patterns/ — hands-on multi-agent orchestration with OpenCode: parallel triage, refactor pipelines, PR-ready CI fixes, doom-loop prevention.
Start at opencode-agent-patterns/README.md — concepts, workflow walkthroughs (sections 0–5), prompts, and cookbook in one file.
| Resource |
Description |
| README.md |
Full guide — primary/subagent model, execution internals, commands, skills, Java examples |
| templates/ |
Copy-paste /triage, /hotfix, refactor, PR-review commands |
| playbooks/ |
Red-build triage, safe refactor, pre-PR review, incident hotfix |
| tools/ |
OpenCode cheat sheet, ecosystem links, example opencode.json agents |
Practice on templates/sample-project/ — includes an intentional failing test for triage exercises.
Complements 12-agents-subagents.md (assistant-agnostic concepts) with OpenCode-specific workflows.