Boundary note: this is a public validation asset, not production deployment and not evidence of a deployed customer workflow.
Incident intake
- What incident or investigation question entered the workflow?
- What severity, service, region, or owner context is available?
- What evidence is missing or uncertain?
- What decision must be made before work proceeds?
Candidate lanes
- Human incident commander or reviewer.
- Runbook or tool-assisted investigation.
- AI-assisted summary or hypothesis drafting.
- Hold, escalate, or narrow scope.
Routing decision
- Which lane was selected?
- Why was that lane safer or clearer than the alternatives?
- What uncertainty remains?
- What context carries forward?
Safe delegation boundary
- What can be summarized, gathered, or drafted?
- What is blocked until a human reviews it?
- What production-impacting or customer-impacting action is out of scope?
- What evidence must not be expanded without approval?
Human review / escalation
- Who reviews the route or scope expansion?
- What condition triggers escalation?
- What approval is required before risky action?
- How is reviewer judgment recorded?
Receipt / replay
- Does the receipt show intake, route, boundary, checkpoint, and final state?
- Are blocked actions visible?
- Can another SRE replay why the investigation stayed human-gated?
- Is the result reviewable without claiming the system handled the incident?
Boundaries
- This is not autonomous incident response.
- It does not claim live incident handling.
- It does not replace SRE ownership.
- It does not prove production operation.