Architecture review

Overview

Use this skill to pressure test structural choices before they quietly become long-term architecture.

When to Use

• new cross-module design
• new dependency or adapter
• structural refactor

When NOT to Use

• purely local stylistic changes
• problems that still need basic framing first

Core Moves

1. State the structural decision.
2. Identify boundaries and dependencies.
3. Compare with a simpler alternative.
4. Name the dominant maintenance risk.
5. Recommend proceed, reduce, or escalate.

Optional Modules

• Complexity budget — Estimate whether the added structure earns its cost. For a feature, make the permanent cost and reversibility explicit (reversible / partially_reversible / one_way_door, with mechanisms like flag, cohort, rollback, migration plan). Delegate to feature-complexity-audit for the full four-dimension scorecard; a one_way_door requires a technical gate, and a high cost approved to build requires a recorded exception_approval on the Feature Value Governance Contract.
• Operational impact — Check deployment, ownership, or observability consequences.
• Future change test — Imagine the next two likely changes and whether the design helps or hurts.
• Module depth review — Evaluate whether a module provides leverage through a smaller interface or merely moves complexity behind a new name. A deep module hides real implementation complexity behind a simple interface; a shallow one just relocates it. Assess interface size versus behavior hidden, a deletion test (what complexity returns if the module is removed), locality (are changes concentrated or scattered), adapter reality (is the seam real or hypothetical), and test surface (is the interface a good place to verify behavior).

Activation Triggers

• Use complexity budget when the proposal adds new layers or abstractions; make reversibility explicit and route high/irreversible cost to a technical gate or exception_approval.
• Use operational impact when runtime behavior or ownership will change.
• Use future change test when the decision could become a template.
• Use module depth review when an abstraction is introduced primarily for naming or layering rather than for proven leverage.

Expected Output

• structural decision summary
• trade-off comparison
• recommendation with main risk

Verification

• The boundary change is explicit.
• At least one simpler alternative was considered.
• Maintenance cost is visible, not implicit.
• For any new module, interface size is weighed against the behavior hidden behind it.
• A deletion test was run: the complexity that returns if the module is removed is named.
• The seam is real, not hypothetical, and the interface is a usable place to verify behavior.

Handoff Signals

• The proposal needs product, security, or platform review.
• The chosen option affects ownership beyond the current team.

Pairs Well With

• feature-planning
• api-design
• triad-check

Anti-patterns

• Treating abstraction as proof of quality.
• Skipping alternatives because the first idea works.
• Focusing on elegance without naming operational cost.