In the previous post, I shared about OpenCode works. Now, let’s go one level deeper. This is the exact workflow I use daily to ship code faster, combining GitHub issue, Copilot Agent, and multiple AI reviewers into a single pipeline.
The Big Idea
AI is not a single tool. It’s a system of agents working together:
-
One writes code
-
Others review
-
Others validate
-
You orchestrate
Step 1 — Start with a Real Issue
Every task begins with a clearly defined issue:
👉 https://github.com/jellydn/my-ai-tools/issues/174
This is critical because:
-
AI needs context
-
Reviews need scope
-
You need a single source of truth
A vague issue = weak AI output.
Step 2 — Let Copilot Do the First Draft
Instead of writing code from scratch, I assign the issue to Copilot then in will open PR and work on it:
👉 https://github.com/jellydn/my-ai-tools/pull/175
At this stage:
-
Copilot generates the implementation
-
I review direction, not syntax
Step 3 — AI Code Review with CodeRabbit
Before touching anything manually, I run an AI review.
CodeRabbit helps:
-
Catch bugs early
-
Identify bad patterns
-
Suggest improvements
This is your first quality gate.
Step 4 — Pull PR Locally
Next, I check out the PR locally using:
This makes it trivial to clone repository and manage under src/tries folder:
-
Checkout PRs
-
Run the code
-
Validate behavior
No friction = more testing.
Step 5 — Multi-Agent Review (This Is the Game Changer)
Now comes the most powerful part. I don’t rely on a single AI. I run multiple agents:
-
Codex
-
OpenCode
-
Pi
-
Claude Code
Each one:
-
Thinks differently
-
Finds different issues
-
Suggests different improvements
This creates a diverse review system, similar to having multiple senior engineers.
Final Thoughts
This workflow helps me:
-
Ship faster
-
Maintain high quality
-
Reduce mental load
The key is not better prompts. It’s better orchestration.