Skip to content

NOTES.md

Latest commit

 

History

History
46 lines (34 loc) · 1.51 KB

NOTES.md

File metadata and controls

46 lines (34 loc) · 1.51 KB

Engineering Team Agent Architecture

Goal

Create OpenCode agents simulating an engineering team with 5 roles:

  • Product Manager
  • Architect
  • Engineer
  • Engineering Manager
  • QA Engineer

Problem

OpenCode supports custom agents via:

  1. JSON config in opencode.json with agent definitions
  2. Markdown files in .opencode/agents/ or ~/.config/opencode/agents/

Each agent can be primary (main assistant) or subagent (invoked via @mention).

For team simulation with inter-agent communication:

  • Subagents can invoke each other via the Task tool
  • Use permission.task to control which subagents each agent can call
  • Each needs custom system prompt defining their role, skills, and collaboration protocols

See: https://opencode.ai/docs/agents

Questions

1. Scope

Global (all projects) or project-specific?

2. Interaction Model

  • Primary + Subagents: You switch between agents using Tab or @invoke; each has distinct role
  • Orchestrator Pattern: Main agent delegates to subagents; you interact with one primary
  • Team Channel: All subagents active in one session, can @mention each other

3. Collaboration Protocol

  • Should agents have a shared context/thread?
  • Do they need explicit meeting/handoff protocols?
  • How should they escalate disagreements?

4. Skills

Each role could have skills (e.g., pm-roadmapping, qa-test-planning). Define now or iterate?

5. Models

Use same model for all, or different models per role (e.g., faster/cheaper for PM, more capable for Engineer)?