Secure service-level objectives operations

Separate runtime and operator roles for AI support-answer service with 99.5% availability + TTFT SLO. Runtime may only perform the narrow actions that reliability owner gating releases on error budget needs; operators get audited break-glass with TTL. Encode a negative probe that denies the privilege trick related to HTTP 200 fallbacks counted as good while citations empty — vanity availability.

Redact prompts/PII at collection. Secrets enter via a manager or workload identity — never source, fixtures, or exception strings. Incident SLO-VANITY-200-18 should be impossible if these controls hold. Output allowlists and schema checks stay in force on error paths.

Rehearse the rollback or kill switch timed against a clock. Record actor, reason, prior revision/secret/flag, and verification query. Invariant reminder: bad events include malformed upstream + timeout fallbacks; auth 4xx excluded.

Abuse cases unique to this lab include the privilege path implied by HTTP 200 fallbacks counted as good while citations empty — vanity availability. Prove a read-only role cannot mutate. Break-glass tokens expire; leftover tokens fail the drill. Dependency pin/digest story matters when images or models move under you. Document how to rotate the credential that AI support-answer service with 99.5% availability + TTFT SLO uses without a full outage window longer than your dual-run plan. Security evidence is part of ship, not an appendix nobody reads.

List network egress destinations and justify each. Ensure debug endpoints are off by default in the shipping config. Verify that error responses do not echo secrets or raw stack frames to clients. For multi-tenant paths, add a cross-tenant probe fixture. Time the rollback drill twice — once with the author, once with a peer. Store the drill transcript beside the threat notes for the incident id. In this chapter the product is AI support-answer service with 99.5% availability + TTFT SLO, the human stakeholder is reliability owner gating releases on error budget, and the incident id you design against is SLO-VANITY-200-18. Re-state the oracle in your notes — 3% provider timeout for 20m trips fast-burn; monthly budget math matches calc-budget.py — and keep the invariant visible: bad events include malformed upstream + timeout fallbacks; auth 4xx excluded. Track slo:availability:ratio and burn_rate1h as the scoreboard. Surface under change control: POST /v1/support/answer. If you only have forty minutes, finish the fixture for HTTP 200 fallbacks counted as good while citations empty — vanity availability before polishing UI. Promotion language stays ternary: promote, hold, or roll back based on evidence, not hope.