Projects, PRDs, and /run-project
Turn larger goals into a to-do list with acceptance criteria, execute stories with back pressure, and finish with verification.

Execution loop
PRDs define the work, /run-project executes stories, and checks prove the result.

Decide if it is a project
In HQ, a PRD is a to-do list with acceptance criteria: small stories, clear done states, and tests or checks that prove completion.
Use /plan when the outcome needs stories, success criteria, coordination, or a new fresh session to execute cleanly.
Run the project
/run-project selects stories, applies worker workflows, checks output, and updates project state. This is the Ralph Loop at project scale.
Be bossy but naive: name the outcome and constraints clearly, then let HQ choose the implementation path and verify it.


Review and land
Project completion includes review, tests, docs or state updates, and a clear next action.
Run the appropriate checks, summarize changed files, capture remaining work or blockers, and hand off before the next major task.
Keep the chapter executable.
Creates an execution-ready project plan with stories and success criteria.
Executes PRD stories with back-pressure checks, commits, and project state updates.
What to keep in mind.
PRD signal
Use a PRD when the work needs multiple stories, tradeoffs, explicit success criteria, or machine-speed execution. Be clear about the outcome and naive about the how.
Completion signal
A story is complete when implementation, verification, and project state all agree. That is the Human Loop above the Ralph Loop: agency, structured problem solving, and first principles.
