rafifuady/agentic-development ? reverse-engineered prompt

Build me a Claude Code plugin for an agent driven software delivery workflow.

I want it to set up a repo so most of the work is done by specialized agents, but the human still has clear checkpoints before anything important gets approved. It should install a command to initialize a project, create the needed config and supporting files in the target repo, and add the checks, hooks, skills, templates, and agents that keep each role writing only in its own lane.

The plugin should also support one task at a time, a gate review command, and a release verification command. Make sure it can be installed once at the user level and then used in any repo, and that the repo side setup can also be done with a plain script if needed.

I also want the guardrails to be real, meaning the lane checks and validation should catch mistakes early, with CI as the final backstop. If you need current Claude Code plugin docs or anything similar, look them up online.