engineering-patterns
Production engineering pattern library spanning 11 phases — Foundation, Resilience, Performance, Data Integrity, Observability, Deployment, Scalability, Edge Cases, Security Deep, Compliance/DR/FinOps
SKILL.mdplugins/loxihart-engineering/skills/engineering-patterns/SKILL.md
GitHubWhen to Use / When NOT
Use when — auditing a service or system for production readiness, or applying a specific hardening phase to existing code.
Do NOT use when — writing the implementation (use engineering-staff-workflows) or making a quick single-file fix that does not need a phase audit.
Engineering Patterns — 11-Phase Production Library
How to Use This Skill
- Audit mode → Run
/engineering-patterns:audit-patternsto score a codebase against all 11 phases - Apply mode → Run
/engineering-patterns:apply-phase <N>to get actionable patterns for a specific phase - Reference → Load any
reference/phase-NN-*.mdfile for deep-dive on a specific concern
The 11 Phases
| # | Phase | Key Concern |
|---|---|---|
| 01 | Foundation | Code quality, testing, CI/CD baseline |
| 02 | Resilience | Error handling, retries, circuit breakers |
| 03 | Performance | Caching, query optimization, profiling |
| 04 | Data Integrity | Transactions, idempotency, validation |
| 05 | Observability | Logging, tracing, metrics, alerting |
| 06 | Deployment | Blue-green, canary, feature flags, rollback |
| 07 | Scalability | Horizontal scaling, queues, sharding |
| 08 | Edge Cases | Concurrency, timeouts, large payloads |
| 09 | Security Deep | Auth, secrets, injection, supply chain |
| 10 | Compliance / DR / FinOps | Audit logs, BCP, cost controls |
| 11 | AI / HITL | LLM integration, prompt safety, human gates |
Scoring System
Each phase is scored 0–5:
- 0 — Not addressed
- 1 — Acknowledged but not implemented
- 2 — Partial implementation, gaps present
- 3 — Implemented, not tested under load/failure
- 4 — Implemented and validated
- 5 — Implemented, tested, monitored, and documented
Target: Phase 1–6 at ≥ 4 before shipping to production. Phase 7–11 at ≥ 3.
Decision Priority (Applied Across All Phases)
Maintainability > Correctness > Security > Observability > Scalability > Performance > Cleverness
Never sacrifice Security silently. Call out conflicts explicitly.
Activation Rules
- Always state which phase(s) you are addressing
- Load the relevant
reference/phase-NN-*.mdbefore making recommendations - Surface a GAPS section if any phase scoring is below 3
- Flag builder's bias if Claude expands scope beyond the requested phase
FAIL Conditions
Named failure modes for this skill. Each is a hard stop, not a warning.
- Applying a later phase (Scalability, Performance) before Foundation and Resilience hold → scaling broken code faster — the phases are ordered, do not skip the baseline.
- Treating the 11-phase list as exhaustive coverage → a domain-specific abuse vector slips through — the list is a floor, not a ceiling; add what the system needs.
- Marking a phase done off its heading without applying the underlying pattern → checkbox theater — each phase requires the actual pattern implemented and verified.
- Silently trading Security for Performance or Cleverness → violates the decision priority — conflicts are called out explicitly, never resolved silently.
- Returning a production-ready verdict without the Security Deep and DR phases → shipping with unmodeled blast radius — those phases are mandatory before any prod-ready claim.
- Using this skill as the builder → it is a pattern checklist, not an implementation engine — route the actual build to engineering-staff-workflows.
Evals
- ✅ SHOULD TRIGGER — "audit this service for production readiness before we ship."
- ✅ SHOULD TRIGGER — "apply the resilience and observability phases to this code."
- ❌ SHOULD NOT TRIGGER — "write the actual retry logic for me" — route to engineering-staff-workflows.