Add the guardrail the postmortem asked for
The limit, validation, or confirmation the postmortem asked for, refusing the exact input that caused the incident.
The ticket
Add the guardrail named in the postmortem so the triggering action cannot repeat.
Acceptance criteria
- The guardrail refuses the exact input or action from the incident
- Legitimate use is unaffected, proven by a test
- The refusal explains what to do instead
- There is a documented way to override it deliberately
What lands as proof
A test replaying the incident's input and being refused, alongside a test proving normal use still works.
Why teams defer it
- Guardrails feel like distrust of the team, so they get softened in review until they no longer guard anything.
- Defining legitimate use precisely is harder than it sounds, and getting it wrong blocks real work.
Questions
- What does the agent actually change?
- The ticket is scoped to one outcome: add the guardrail named in the postmortem so the triggering action cannot repeat. Work that serves that outcome is in scope, and anything outside it is left for a separate ticket, so the pull request stays reviewable.
- How do I know the work is done?
- The pull request carries the evidence, not only the diff. Here that means the guardrail refuses the exact input that caused the incident, so a reviewer can confirm the result without reproducing the work locally.
- How much oversight does this need?
- The run stops once the plan is written. Somebody reads the approach and approves it before any code exists, which is the cheapest moment to redirect the work.