enhance(feature-dev): add system design patterns, edge cases, and operational context to code-architect agent

[sourabharsh]

Summary

Enhances the code-architect agent in the feature-dev plugin with three new steps that bridge the gap between high-level system design and codebase-specific architecture.

Changes

Step 0: System Design Pattern Analysis (new)
Before examining codebase-specific patterns, identify 2-3 relevant system design patterns (event-driven, CQRS, circuit breaker, cache-aside, saga, etc.) and describe how each applies to the feature requirements, expected load profile, and failure modes. Grounded in established industry references (Google SRE, AWS Well-Architected Framework, microservices.io).

Step 2: Edge Cases & Test Expectations (new)
Enumerate 10-15 edge cases with expected behavior and performance characteristics. Each edge case is mapped to a system design pattern from Step 0. Prioritized by marginal return of utility: focus on cases where additional handling yields high value relative to implementation cost, and flag cases where further handling yields diminishing returns.

Step 5: Operational Context (new)
Specify key logging points, metrics to emit, alerting thresholds, and applicable SLOs (e.g., p99 latency, error rate budgets). References industry-standard observability patterns.

Step 3: Architecture Design (enhanced)
Now evaluates the chosen approach against system design patterns and edge cases from Steps 0 and 2. Uses marginal return of utility as a decision framework: prefer the approach that maximizes value per line of code and operational complexity.

Output Guidance (updated)
Three new sections added: System Design Patterns, Edge Cases & Test Expectations, and Operational Context.

Motivation

The current code-architect agent jumps directly from codebase pattern analysis to architecture design. This means:

1. No system-level design context before diving into codebase-specific patterns
2. No proactive edge case enumeration (Phase 3 of the feature-dev workflow mentions edge cases but only as questions to ask the user, not as architect deliverables)
3. No operational context (logging, metrics, SLOs) in the architecture blueprint

These additions align with the existing feature-dev philosophy of "understand before acting" and "design thoughtfully before implementing" but extend it to system design and production readiness.

Test plan

• Verify code-architect agent triggers correctly with /feature-dev command
• Confirm agent output includes all new sections (System Design Patterns, Edge Cases, Operational Context)
• Test with a simple feature to ensure the agent does not over-enumerate edge cases for trivial tasks
• Verify marginal return of utility framing does not add excessive verbosity to output