A reusable multi-agent orchestration prompt for coding workflows and builds

doodlestein · x · 2026-07-25

The post shares a large orchestration prompt for users of the author’s skills tooling ecosystem.

It tells the agent to first read all of AGENTS.md and README.md, then fully understand the codebase, architecture, and project purpose.
It then instructs the system to use code-investigation mode plus a multi-agent swarm: 3 coding instances and 3 Claude Opus instances, both at high thinking effort.
The prompt also emphasizes operational hygiene: periodically clear stale build artifacts, avoid running out of space, and use the rch skill for all builds and tests.
The core value here is a reusable workflow prompt for coordinating multiple coding agents and keeping build/test execution disciplined.

More from coding & agent