Plan Enterprise

Skill: Plan Enterprise

Purpose

Orchestrate the full METHOD-ENTERPRISE-BUILDER-PLANNING 8-phase cycle for enterprise-grade, mission-critical, and high-availability software systems.

When to use

Activate this skill when the user requests:

• "Plan an enterprise system..."
• "Design a mission-critical [component]..."
• "Build a high-availability [module] with [SLA]..."
• "Architect a security-first [system] with [compliance]..."

Or when the user mentions regulated data (PCI-DSS, GDPR, HIPAA, SOC2, ISO 27001).

Workflow

Phase 1: Enterprise Context Analysis

Actions:

1. Classify system (enterprise-grade, mission-critical, high-availability, security-first, ACID-required)
2. Build stakeholder map (end users, compliance officer, operations, regulators)
3. Map regulatory environment (applicable standards with obligations)
4. List integration points (protocol, SLA dependency level)

Output artifact:

• Create docs/enterprise-context.md (use createFile operation)
• Artifact type: implementation_plan

Gate: Ask user to confirm system classification before proceeding.

Phase 2: Non-Functional Requirements

Actions:

1. Define Performance SLOs: throughput (TPS), latency p95/p99
2. Define Availability SLA: % uptime, RTO, RPO, deploy strategy
3. Define Scalability: initial, peak, growth rate
4. Define Security baseline: auth, authz, encryption at rest/transit
5. Define Compliance per standard

Gate: All SLOs must be numeric and measurable (not "fast" or "secure" — "p95 ≤ 500ms", "99.9% SLA").

Phase 3: Risk Matrix

Actions:

1. STRIDE analysis on primary API surface (Spoofing, Tampering, Repudiation, Info Disclosure, DoS, Elevation of Privilege)
2. Technical risk catalog (probability, impact, mitigation)
3. Flag any CRITICAL unmitigated risk — blocks Phase 4

Output: docs/risk-matrix.md

Gate: No CRITICAL risk may be unmitigated; all HIGH risks must have mitigations.

Phase 4: Micro-task Decomposition

Actions:

1. Identify business domains
2. Map domains to layers (infrastructure → domain → application → interface)
3. Create one micro-task per file/unit, each ≤50 effective lines
4. Assign task IDs (DOMAIN-LAYER-NNN)
5. Define dependency order (topological sort)

Validation per micro-task: After implementing each task, load @skill lint-microtask to validate effective line count.

Output: Backlog table in Phase 4 section

Phase 5: Architecture Decisions (ADR)

Actions:

1. Apply architecture decision tree
2. Create one ADR per significant decision
3. Each ADR must have: Context, Decision, ≥2 Alternatives with rejection reasons, Consequences, Compliance Impact
4. Generate C4 diagrams (L1 System Context + L2 Container)

Validation: After generating each ADR, load @skill validate-adr to check against 11 enterprise rules.

Output: docs/adr/ADR-NNN-title.md (one file per decision)

Gate: ADR cannot be marked Accepted if validation fails.

Phase 6: Security & Compliance Mapping

Actions:

1. Define ACID transaction boundaries for all write operations
2. Map each compliance requirement to a concrete control
3. Define RBAC roles and ABAC ownership rules
4. Define audit event schema (what, where, retention)

Phase 7: Test Strategy

Actions:

1. Define test pyramid allocation (Unit 70%, Integration 20%, E2E 7%, Chaos 3%)
2. List Critical User Journeys for E2E
3. Define load test scenarios (baseline, spike, soak) with numeric pass criteria
4. Define CI/CD quality gates

Coverage requirement: ≥99% line + branch (floor — assertions must verify behavior).

Phase 8: Delivery Report

Actions: Generate delivery report with:

• Micro-tasks completed/total
• Test results (unit/integration/E2E/load)
• Coverage % per layer
• STRIDE status
• Critical/High vulnerabilities: 0
• ADR status
• Compliance matrix
• STATUS: READY FOR PRODUCTION or BLOCKED

Artifact type: delivery_report

Progressive skill loading

Do not load all phase skills at once. Load only the skill for the current phase:

Phase 1 in progress → load @skill phase-1-context
Phase 5 in progress → load @skill phase-5-adr
Validation needed   → load @skill validate-adr or @skill lint-microtask

This prevents context saturation while maintaining full capability.