Review Domain Boundaries

Quick Reference

HARD-GATE

DO NOT recommend splitting code into new contexts unless the business boundary is explicit enough to name.
DO NOT treat every large module as a bounded context automatically.
ALWAYS identify the leaked language or ownership conflict before proposing structural changes.

Core Process

Core principle: Fix context leakage before adding more patterns.

When to Use

• Next step: Chain to model-domain when a context is clear enough to model tactically, or to refactor-code when boundaries need incremental extraction.

Review Order

1. Map entry points: Start from controllers, jobs, services, APIs, and UI flows that expose business behavior.
2. Name the contexts: Group flows and rules by business capability, not by current folder names alone.
3. Find leakage: Look for terms, validations, workflows, or side effects crossing context boundaries.
4. Check ownership: Decide which context should own invariants, transitions, and external side effects.
   • For Fleet/Billing examples, Billing owns invoice generation triggers and invoice side effects; Fleet owns vehicle state and availability. Flag Fleet::Vehicle triggering invoice generation as leakage into Billing, not the reverse.
5. Propose the smallest credible improvement: Rename, extract, isolate, or wrap before attempting large reorganizations.

Detecting Leakage

Use ripgrep to find cross-context references before reading code manually:

# Find references from one context into another
rg 'Billing.*Fleet|Fleet.*Billing' app/

# Find cross-namespace constant usage
rg 'Billing::[A-Z]' app/services/fleet/
rg 'Fleet::[A-Z]' app/services/billing/

# Find callbacks that touch foreign concepts
rg 'after_(create|update|save).*Job|after_(create|update|save).*Mailer' app/models/

# Find invoice-generation triggers leaking out of Billing
rg 'invoice|Invoice|Billing' app/models app/services app/jobs

Common Pitfalls

• Treating a shared database table as proof of a shared context — storage and domain boundaries are independent concerns.
• Splitting into new contexts before the business language is stable enough to name them clearly.
• Mistaking a large Rails namespace for a bounded context without checking whether it has a single, coherent set of rules and an identifiable owner.

Output Style

1. Finding Format: For each finding include:
   • Severity
   • Contexts involved
   • Leaked term / ownership conflict
   • Why the current boundary is risky
   • Smallest credible improvement Include the ownership direction when it matters (e.g. Billing should own invoice-generation triggers; Fleet should not trigger invoices from Fleet::Vehicle).
2. Structure: Write findings first, then list open questions and recommended next skills.
3. Language: Must be in English unless explicitly requested otherwise.

Extended Resources

Load only when a concrete boundary-leakage example is needed:

• EXAMPLES.md — Billing/Fleet leakage example with smallest credible fix.

Integration