Agent Skills: Pattern Enforcement

Pattern Enforcement

Critical rules

• Patterns survive only as 'error' lint/type/CI failures — not README suggestions.
• Enforce layer boundaries (domain must not import infra), object params, and server/client separation.
• Compiler-level any/as bans belong in strict-typescript; this skill enforces structure.
• Adopt incrementally if needed, but keep rules on so new violations cannot land.
• Before editing ESLint config, read references/eslint-config.md.

Workflow

1. Map layers (domain / infra / api) and client vs server globs.
2. Add boundaries (or no-restricted-imports), prefer-object-params, and no-server-imports / no-nodejs-modules.
3. Set every architectural rule to 'error'. Wire ESLint into CI so failures block merge.
4. For existing debt, follow references/migration.md (warn → fix → error, or codemod).
5. Verify agents and humans get deterministic failures on pattern drift.

Resources

• references/eslint-config.md — boundary/object-param/server rules and starter config. Read when configuring.
• references/migration.md — incremental adoption, AI rationale, rationalizations. Read when remediating.

Validation

• [ ] Directional boundary rules defined and enforced
• [ ] prefer-object-params on (off only in tests)
• [ ] Server/client separation enforced on client globs
• [ ] Architectural rules are 'error'; CI fails on ESLint errors
• [ ] Existing violations have a tracked migration plan

Constraints

• Does not replace teaching a single pattern in depth — load that skill. Does not enforce runtime values (use validation-boundary).
• Related: strict-typescript, fn-args-deps, config-management, api-design.