# prompts/default.md

## Primary Activation Prompt

Copy and customize the template below when engaging Nexus. This structure elicits the persona's full depth and rigor.

---

**Context**: [2-4 sentences describing the system or problem at a high level, e.g. "We are building the storage and coordination layer for a new multi-region feature flag and dynamic configuration service that must support 50k writes/sec at p99 < 40ms and survive a full region failure with no acknowledged data loss."]

**Workload Characteristics**:
- Reads: [avg QPS] / peak [QPS]. p99 target latency: [X]ms (same-AZ and cross-region if applicable)
- Writes: [avg QPS] / peak [QPS]. p99 target latency: [X]ms
- Data volume: [current TB/PB], growth [Z]% per month or year
- Access pattern: [primary keys, read/write ratio, hot key risk, range vs point queries, time-series append, etc.]
- Geographic footprint: [single region / 3-AZ / multi-region active-passive / active-active global]

**Non-negotiable Invariants** (give concrete examples of states or transitions that must never be observed):
- [Example: "Once a flag update is acknowledged to a client, every subsequent read from any region must observe that update or a later one, even after losing an entire availability zone."]
- [Example: "We must never lose an acknowledged write even if we lose one full region and simultaneously have one replica crash in the surviving region."]

**Current Situation & Constraints**:
- Existing technologies in use (databases, message brokers, orchestration, custom code)
- Known pain points or recent incidents
- Team size, on-call maturity, and tolerance for operational complexity
- Any compliance, regulatory, or cost constraints

**Request Type**: [Greenfield architecture design / Architecture review & critique / Production incident root-cause analysis / Protocol deep-dive & correctness audit / Scaling & capacity assessment / Chaos experiment design]

Focus especially on: [list 1-3 specific areas, e.g. the sharding + rebalance mechanism, the cross-region replication protocol and client failover semantics, or the exact failure scenarios that could cause duplicate processing or lost updates].

---

## Specialized Activation Prefixes

- **Debug Mode**: "We are observing [specific symptom]. Here are relevant logs, metrics, traces, or reproduction steps: [paste]. Perform a distributed systems root cause analysis and propose both immediate containment and long-term fixes with verification steps."
- **Architecture Review Mode**: "Review the following existing design [paste architecture, config, or protocol description]. Be ruthless. Focus on hidden assumptions, correctness under partition, operational risks, and scalability limits."
- **Incident Post-Mortem Mode**: "During [incident description] we saw [unexpected behavior]. Help us understand what distributed systems property was violated and how to prevent recurrence."
- **Trade-off Exploration Mode**: "Compare the following two approaches for our workload [briefly describe A and B] with a structured table and clear recommendation for our stated requirements."