| name | launch-readiness-review |
| description | Run a go/no-go launch-readiness review against written, pre-agreed criteria and design a staged rollout with a tested rollback. Use as a launch approaches — define measurable, owner-assigned criteria first, then facilitate the decision and record any waiver with its risk acceptance. |
Skill: Launch-readiness review
Turn "we think we're ready" into a documented go/no-go and a staged rollout. A
launch decision made on vibe is an accident waiting for a retro.
When to use
- A program milestone or product is approaching go-live.
- A team wants to "just ship it" with no agreed bar for ready.
- A big-bang launch is planned with no rollout stages or rollback.
Procedure
- Define criteria before the review. Each go/no-go criterion is measurable and
owner-assigned, agreed in advance. Use the
launch-readiness-checklist
template. Criteria invented in the room are bias, not a bar.
- Facilitate the go/no-go. Walk each criterion: GO / NO-GO / waived. Traverse
the go/no-go tree in
../../knowledge/tpm-engagement-decision-trees.md.
- Reach an explicit decision with an owner. Record who made the call. A waived
criterion is recorded with its risk acceptance and who accepted it — never
silently dropped.
- Design the staged rollout. Canary → percentage ramp → full, each stage gated
on a specific metric (error rate / latency / conversion) that must hold to
advance.
- Test the rollback. Name the rollback path and who can pull it, and verify it
works before go-live. An untested rollback is not a plan.
Output
A completed readiness checklist, a recorded go/no-go decision (with any waiver +
risk acceptance), and a staged rollout + tested rollback plan.
Anti-patterns (the hook flags these)
- A launch checklist with no go/no-go criteria or pass/fail bar.
- A "we'll watch it closely" gate (no metric, no threshold).
- A rollout with no stages or an untested rollback.