Problem
The first live project will hit conditions no test covers: a contributor stalls, the gate
holds something unexpectedly, CI breaks on the scaffold, the customer asks for a change
mid-flight. None of that has a procedure.
Scope
- Runbook covering: intake through delivery, what the operator does at each stage, and what
is automated versus manual in v1
- Failure handling: stalled issue, repeatedly rejected PR, contributor abandonment, CI
failure on the default branch, criteria that turn out to be unsatisfiable
- Customer communication points — what they see and when, and what requires their input
- Escalation: what stops the loop, who is paged, and what the customer is told
- Explicit statement of v1 limits: what we do not accept as a project, and what we do not
promise
Acceptance
Someone other than the author could run a project from brief to delivery using only the
runbook.
Problem
The first live project will hit conditions no test covers: a contributor stalls, the gate
holds something unexpectedly, CI breaks on the scaffold, the customer asks for a change
mid-flight. None of that has a procedure.
Scope
is automated versus manual in v1
failure on the default branch, criteria that turn out to be unsatisfiable
promise
Acceptance
Someone other than the author could run a project from brief to delivery using only the
runbook.