Single-Shot Multi-Agent SDLC with Loop & Graph Engineering

What can be the way to single shot this implementation like the way Anthropic engineering does under karpathy ? Assuming we are using multi-agent development systems, using loop and graph engineering 
https://www.towardsdeeplearning.com/andrej-karpathy-asked-for-a-tool-48-hours-later-graphify-went-viral-10d8ead5f50e

https://github.com/cobusgreyling/loop-engineering 

...

Essentially my goal is this 
Project Manager - oversees all agents and their progress. Acts as a arbitrator and final decision maker for go no go across all milestones. Primarily AGENTS.md file kind keeping track of who needs to do what.
Coder - Coding agents perform individual unit task
DBA - Responsible for database communications, schema management and integrity, master data etc.
QA - Responsible for testing individual units and telling coders what to fix it before giving a sign off on units.
Team lead - Performs code merge of distinct units and coordinates with the QA for automated integrated testing
Architect - Creates DRY principles and architectural design spec, audits code where DRY principles are not followed and hands of the report to Project Manager. PM prioritizes what needs to be fixed and what can be delegated later phases. 
DevOps - Manages local git health, performs db connectivity checks and keeps a track of environment variables and other artifacts. 

---

Based on the above suggest me create master prompt to execute this entire project in a single shot development. 

I already have a operator code here, copy it to the local path and use this operator in a new port so that human can verify the progress of the implementation across all sprints in the dashboard / operator console. 

/mypath/operator

--

Make sure the operator is console is started first followed by master prompt execution and progress updates by the agents.