nw-ab-critique-dimensi…
Review dimensions for validating agent quality - template compliance, safety, testing, and priority validation
Hexagonal architecture patterns with pure core and side-effect shell for functional codebases
$ npx -y skills add nWave-ai/nWave --skill nw-fp-hexagonal-architecture --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/nw-fp-hexagonal-architectureContext preview
The summary Claude sees to decide when to auto-load this skill.
Hexagonal architecture patterns with pure core and side-effect shell for functional codebases
name: nw-fp-hexagonal-architecture agent: nw-functional-software-crafter description: Hexagonal architecture patterns with pure core and side-effect shell for functional codebases user-invocable: false disable-model-invocation: true
Ports and adapters in functional programming. Structure applications with a pure core and side-effect shell.
Cross-references: [fp-principles](../nw-fp-principles/SKILL.md) | [fp-domain-modeling](../nw-fp-domain-modeling/SKILL.md) | [fp-usable-design](../nw-fp-usable-design/SKILL.md)
---
[STARTER]
Functional architecture naturally implements ports and adapters. The paradigm's separation of pure functions from side effects IS the hexagonal boundary.
| OOP Concept | FP Equivalent | Why | |---|---|---| | Port (interface) | Function type signature / type alias | Port defines contract; function signature IS that contract | | Adapter (class) | Concrete function implementation | Adapter fulfills contract; matching function does same | | DI container | Function parameters / partial application | Dependencies passed as arguments, no container needed | | Domain service class | Module of pure functions | Related pure functions replace stateful service object | | Entity with behavior | Immutable data + functions operating on it | Data and behavior separated; functions transform immutable values |
---
[STARTER]
All business logic is pure; all side effects live at the system's edges.
**The Sandwich Pattern**: Read (impure) -> Decide (pure) -> Write (impure)
+--------------------------------------------------+ | Side-Effect Shell (thin) | | - HTTP handlers, CLI, message consumers | | - Database access, file I/O, network calls | | - Reads data, calls core, writes results | | | | +--------------------------------------------+ | | | Pure Core (large) | | | | - Pure functions only | | | | - Domain logic, validation, calculation | | | | - No I/O, no side effects | | | | - Immutable data transformations | | | +--------------------------------------------+ | +--------------------------------------------------+
**Dependency Rule**: Shell may call core. Core never calls shell. Core is unaware of shell's existence.
**Why**: Pure core is trivially testable (no mocks, no setup, no teardown). Shell is thin and needs few integration tests.
---
[STARTER]
A port is a function type signature describing a capability the domain needs:
FindOrder : OrderId -> AsyncResult<Order option> SaveOrder : Order -> AsyncResult<unit> SendEmail : Email -> AsyncResult<unit> GetPrice : ProductCode -> Price CheckExists : ProductCode -> bool
**When to define**: Domain needs a capability involving I/O or external systems. Domain declares WHAT; adapter provides HOW.
**Naming**: Verb-noun. Name describes capability, not technology.
---
[STARTER]
An adapter is a concrete function matching a port's type signature:
PostgresOrderRepo.findOrder : OrderId -> AsyncResult<Order option> InMemoryOrderRepo.findOrder : OrderId -> AsyncResult<Order option>
Both match the `FindOrder` port. Domain doesn't know which is used.
---
[STARTER] -> [INTERMEDIATE] -> [ADVANCED]
How many dependencies does the function need?
1-3 --> [STARTER] Functions as Parameters
4-6 --> [INTERMEDIATE] Consider Environment Pattern or grouping
7+ --> [ADVANCED] Capability Interfaces or Effect System
(also: reconsider function responsibilities)Pass dependencies as function parameters. Partially apply at composition root.
placeOrder (findCustomer) (saveOrder) (rawOrder) = ... placeOrderHandler = placeOrder Database.findCustomer Database.saveOrder
Dependencies in a record, provided once at top level. Use when parameter threading becomes painful (4+ deps).
placeOrder (rawOrder) = reader { env = ask(); env.findCustomer(rawOrder.customerId) ... }
placeOrder(rawOrder) |> runWith(productionEnv)Abstract over effect types (tagless final) or use fine-grained effect tracking (ZIO, Koka). Use for large codebases with many effects.
| Context | Approach | |---|---| | Small/medium codebase | Functions as parameters | | Large codebase, many effects | Capability interfaces or effect system | | Pragmatic TypeScript/F# | Functions as parameters + modules |
---
[INTERMEDIATE]
Workflows flow through architecture as pipelines:
HTTP Request -> Parse (shell: impure) -> Validate (core: pure) -> Calculate (core: pure) -> Persist (shell: impure) -> Respond (shell: impure)
Each pure step is a function in the pipeline. Shell handles I/O at start and end.
**Error-track pipelines**: Each step returns Result type; pipeline short-circuits on first failure. See [fp-domain-modeling](../nw-fp-domain-modeling/SKILL.md).
**Collect-all-errors**: When you need ALL validation errors, use Applicative style. See [fp-principles](../nw-fp-principles/SKILL.md) section 5.
---
[INTERMEDIATE]
| Layer | Test Type | Volume | Speed | Mocks | |---|---|---|---|---| | Pure core (domain) | Unit + Property-based | Many | Fast (ms) | None | | Composition root | Integration (wiring) | Few | Medium | None | | Adapters | Integration | Few per adapter | Slow | None (real deps) | | End-to-end | System tests | Very few | Slowest | None |
**Key insight**: Pure functions need no mocking. Input in, outpu
AI agents that guide you from idea to working code, with human judgment at every gate. nWave runs inside Claude Code. It breaks feature delivery into seven waves (discover, diverge, discuss, design, devops, distill, deliver).
Repo: nWave-ai/nWave
Review dimensions for validating agent quality - template compliance, safety, testing, and priority validation
Review dimensions for validating agent quality - template compliance, safety, testing, and priority validation
Review dimensions for acceptance test quality - happy path bias, GWT compliance, business language purity, coverage completeness, walking skeleton…
Detailed 5-phase workflow for creating agents - from requirements analysis through validation and iterative refinement
5-layer testing approach for agent validation including adversarial testing, security validation, and prompt injection resistance
Architectural style selection decision matrices, trade-off analysis, structural enforcement rules, and combination patterns. Load when choosing or evaluating…