## 🎯 Default Prompt Template — Senior Distributed Systems Engagement

Copy, fill in the brackets, and send to activate Nexus at full strength.

---

### Context

I need senior distributed systems help on: **[architecture design | design review | incident analysis | data pipeline | multi-region | API/platform reliability]**

**System one-liner:** [What it does in one sentence]

**Current stage:** [greenfield | pre-launch | scaling pain | post-incident hardening]

### Workload & Requirements

- **Traffic shape:** ~[QPS/RPS], read:write ≈ [ratio], payload ~[size], peak multiplier [Nx]
- **Latency targets:** p50 [ ], p99 [ ], hard deadline? [yes/no]
- **Availability / RTO / RPO:** [e.g., 99.9%, RTO 15m, RPO 0/minutes]
- **Consistency needs:** [strong | read-your-writes | eventual | per-entity serializable | other]
- **Multi-tenancy:** [no | yes—describe noisy-neighbor risks]
- **Compliance / data residency:** [none | GDPR | region pin | other]

### Constraints

- **Stack (if fixed):** [languages, DBs, queues, cloud]
- **Team / ops reality:** [on-call maturity, headcount, existing runbooks]
- **Non-goals:** [what we are explicitly not solving]
- **Deadline:** [when a decision is needed]

### Artifacts Attached / Described

- Current design: [paste RFC summary, diagram-as-text, or “none—greenfield”]
- Known pain: [symptoms, metrics, recent incidents]
- Prior decisions: [what was already tried or rejected]

### Ask

Please provide:

1. **TL;DR recommendation**
2. **Problem framing & assumptions** (call out anything you inferred)
3. **2–3 design options** with trade-offs (consistency, latency, cost, ops complexity, failure modes)
4. **Your recommended path** + why + reverse triggers
5. **End-to-end failure analysis** (partition, node death, slow dependency, poison message, skew)
6. **Observability & SLO sketch** (SLIs, alerts, traces/logs that matter)
7. **Rollout / migration plan** with canary and rollback
8. **Top risks & open questions**

Be opinionated. Prefer boring, operable designs. Flag anti-patterns sharply.

---

### Optional Variants

**Incident mode:** Replace “Ask” with: stabilize → scope → hypotheses → evidence plan → mitigations → RCA structure → permanent fixes.

**RFC review mode:** Review the attached design against correctness, reliability, operability, cost, and migration risk. Separate P0 blockers from nits.

**Deep dive mode:** Teach the core mechanism (e.g., Raft fencing, outbox pattern, quorum) with a worked example and production pitfalls.