# Guardrail scope decision record

**Proposed blank worksheet · after a validated fix · one record per protected failure**

Decision/status: [proposed / evaluating / active / retiring]  
Owner: [named person] · Tracking item: [link] · Review date: [date]

| Field | Complete before adoption |
|---|---|
| **Failure and check** | [Trigger; faulty/corrected behavior; user effect; reproduction evidence and conditions; proposed protection; detection/acceptance evidence; coverage limits.] |
| **Protected contract** | [Workflow start/end and completion promise; eligible population; correctness requirements, including relevant failure/state cases; exclusions.] |
| **Passing boundary** | [Allowed behavior or limit; workflow/contract rationale; supporting observations and variation; sample/window if relevant; failures/incomplete attempts; inconclusive rule.] |
| **Chosen scope** | [Local / team / platform; component(s) and consumers; mechanism/contract compatibility evidence; evidence missing for wider reuse.] |
| **Response and rollout** | [Block / review / ticket / page; recipient/triage; initial exposure; expansion and stop conditions; rollback where relevant; temporary bypass owner, reason and expiry.] |
| **Cost and false alarms** | [Execution, environment, fixture, triage and support cost; false-alarm evidence/review method; acceptable upkeep; recurring noise response.] |
| **Ownership and review** | [Named maintainer; consumer responsibilities; tracking item; review date and event triggers.] |
| **Retirement and decision** | [Removal/replacement criteria; replacement coverage evidence; remaining risk; adopt/revise/defer/gather evidence/retire; next action and owner.] |

**Decision:** [Adopt at the stated scope / gather evidence / revise / defer / retire. Explain why.]  
**Next action and owner:** [Concrete action, named owner, completion evidence and due date.]

Wider reuse requires demonstrated fit across consumers. Recurrence alone does not justify platform scope. Leave a numeric boundary blank when data and variation do not justify one.

This worksheet is proposed. It does not report a completed experiment or guardrails implemented in the author’s published cases.