Aoa Engineering Shape

aoa-engineering-shape

Intent

Choose one shaping lens and return the smallest target-bound responsibility boundary without installing an architecture framework by default.

Trigger boundary

Use this skill when:

• domain responsibilities are overloaded, stable rules are mixed with glue, or a concrete dependency leaks across a meaningful seam

Do not use this skill when:

• the real task is source authority, decision history, verification, an ordinary edit, or implementation after the boundary is already explicit

Inputs

• target system or code slice, target-specific owner surfaces, current responsibilities
• concrete ambiguity, repeated rule, or leaking dependency

Outputs

• exactly one typed mode result with boundary, interfaces, owner or unresolved owner edge, migration edge, verification need, unresolved question, and stop line

Procedure

1. Read references/contract.yaml and choose exactly one mode:

   Mode	Select when	Required procedure
   contexts	Meanings, responsibilities, owners, or dual postures are overloaded.	references/contexts.md
   core	A stable rule is mixed with loading, rendering, orchestration, or infrastructure.	references/core.md
   port-adapter	A concrete database, API, filesystem, CLI, provider, or runtime detail leaks inward.	references/port-adapter.md

2. Read the selected reference completely. Do not load the other mode procedures merely because their vocabulary appears nearby.

3. Bind owner only from a declaration that governs the target slice. An unrelated authored surface, sibling route, recent file, or generated index cannot supply the target owner. Keep owner: unresolved when that binding is missing; never borrow a plausible owner.

4. Execute only the selected procedure. If another shaping mode is materially necessary, finish the upstream result and hand its typed output to a later task-local DAG node instead of blending modes.

Contracts

• owner facts remain with their owner; this bundle supplies procedure only
• proposed boundaries may continue with an unresolved owner, but cannot claim adoption, placement authority, or an owner-specific destination
• one primary responsibility center; cross-relations stay explicit
• errors, fallback, and limits are part of a port contract, not hidden adapter behavior
• a proposed boundary is not accepted architecture until its owner adopts it

Risks and anti-patterns

• generic clean-architecture diagrams without exact local responsibilities
• assigning a target owner from a nearby but unrelated authoritative file
• making directories, frameworks, or current technique names the semantic core
• creating a port with no consumer or a context map for a spelling fix

Verification

• confirm the selected mode matches the actual pressure
• cite the target-specific responsibility/rule/dependency and the exact owner binding, or state that the owner remains unresolved
• state migration compatibility, what remains unresolved, and what this result does not authorize

Stop after one reviewable boundary and next-owner question. Do not implement or multiply abstractions unless the request separately authorizes that effect.

Adaptation points

Owners provide local vocabulary, interfaces, compatibility windows, and validation commands. Those facts remain owner data, not forks of this skill.