Agent Skills: The Functional Core, Imperative Shell Paradigm

The Functional Core, Imperative Shell Paradigm

When To Use

• Separating pure business logic from side effects
• Improving testability through immutable domain models

When NOT To Use

• Performance-critical hot paths where immutability overhead matters
• Purely imperative codebases with no plans to adopt functional patterns

When to Employ This Paradigm

• When business logic is entangled with I/O operations (e.g., database calls, HTTP requests), making tests brittle and slow.
• When significant development time is spent rewriting adapters or dealing with framework churn.
• When you require a suite of fast, deterministic unit tests that operate on plain data, complemented by a thin integration testing layer.

Adoption Steps

1. Inventory Side Effects: Create a map of all side effects in the system, such as database writes, external API calls, UI events, and filesystem access. Explicitly assign these responsibilities to the "shell."
2. Model the Core Logic: Represent business rules and policies as pure functions. These functions should take domain data as input and return decisions or commands as output, avoiding shared mutable state.
3. Design the Command Schema: Define a small, explicit set of command objects that the core can return and the shell can interpret (e.g., PersistOrder, PublishEvent, NotifyUser).
4. Refactor Incrementally: Begin with high-churn or critical modules. Wrap legacy imperative code behind adapters while progressively extracting pure calculations into the functional core.
5. Enforce Boundaries: Use code reviews and automated architecture tests to validate a strict separation. The shell should only handle orchestration, sequencing, and retries, while the core should never call directly into frameworks or I/O libraries.

Key Deliverables

• An Architecture Decision Record (ADR) detailing why this pattern was chosen, which modules are affected, and the scope of the migration.
• A suite of unit tests for the core with high (>90%) and deterministic code coverage. Where applicable, use property-based or fixture-based testing to cover a wide range of inputs.
• A suite of contract and integration tests for the shell that verify correct command interpretation, retry logic, and telemetry.
• A set of rollout metrics (e.g., deployment lead time, incident rate in the shell layer) to demonstrate the value of the architectural change.

Risks & Mitigations

• Logic Drifting Between Core and Shell:
  • Mitigation: It's common for business logic to accidentally be duplicated or placed in the shell. Enforce a "core owns all decisions" checklist during code reviews to prevent this.
• Mismatch with Frameworks:
  • Mitigation: The imperative shell may still need to interact with framework-specific lifecycle hooks. Before committing to a large rewrite, build small proof-of-concept adapters to validate the integration strategy.
• Team Unfamiliarity with the Pattern:
  • Mitigation: Introduce the pattern using pair programming and internal "brown-bag" learning sessions. Document common anti-patterns that are discovered during the pilot phase to guide future development.

Concrete Components

These vocabulary items name the concrete tools and abstractions that show up when the paradigm is implemented. They are not required dependencies and they are not part of the skill's tools: frontmatter (which is reserved for Claude Code tool restrictions). Use this list to disambiguate during architecture discussions.

• boundary-validator: guards inputs to the pure core so the core can stay total
• core-test-generator: generates property-based tests against the deterministic core
• shell-adapter-generator: scaffolds the imperative shell that wires the core into I/O

Exit Criteria

• [ ] An explicit side-effect inventory exists mapping every I/O operation to the shell layer before any refactoring begins.
• [ ] Core unit tests run with no I/O mocks (no database, HTTP, or filesystem calls) and achieve

  90% deterministic coverage.

• [ ] Automated architecture tests (e.g., import analysis or namespace rules) confirm the core contains zero imports from I/O or framework libraries.
• [ ] Rollout metrics (deployment lead time, incident rate in the shell layer) are baselined before migration and compared after to demonstrate the pattern's value.