directive

Understand Directive fundamentals: what modules, facts, derivations, constraints, resolvers, and systems are, and how they fit together. Use when someone is new to Directive, asks 'what is Directive', wants to understand the mental model, or needs help choosing between Directive concepts.

Add guardrails (input/output validation), memory strategies, token budgets, PII detection, prompt injection prevention, and circuit breakers to AI orchestrators. Use when enforcing content policies, managing costs, preventing security vulnerabilities, adding resilience, or implementing agent memory across turns.

Compose Directive modules into systems: single-module and multi-module createSystem() calls, namespaced fact/derive/event access, cross-module dependencies, initialFacts and hydration, plugins (logging, devtools, persistence, circuit breaker), React adapter hooks (useSelector, useEvent, useSystem, DirectiveProvider), and system lifecycle. Use when creating a system, connecting modules together, adding plugins, or integrating Directive with React.

Write Directive constraints, resolvers, and error boundaries: when/require patterns, static vs dynamic requirements, async constraints with explicit deps, priority and ordering, resolver retry policies, deduplication keys, batch resolution, and system error boundaries. Use when asked to add constraint-resolver pairs, handle errors/retries in resolvers, or configure recovery strategies.

Configure AI provider runners (Anthropic, OpenAI, Ollama), stream tokens and structured output, and wire cross-agent communication. Use when setting up LLM provider connections, implementing token streaming, choosing between providers, or building agents that communicate results to each other.

Create and structure Directive modules with schema definitions, init functions, derivations, effects, events, constraints, and resolvers using the correct t.* type builders and naming conventions. Use when asked to build a Directive module, add facts/derivations/resolvers to an existing module, or scaffold any stateful domain with Directive.

Migrate state management code from Redux, Zustand, XState, MobX, Recoil, or Jotai to Directive. Provides concept mapping, step-by-step migration patterns, and before/after code examples. Use when asked to migrate, convert, or port existing state management to Directive.

Review Directive code for anti-patterns, naming violations, missing error boundaries, constraint/resolver misuse, and performance issues. Use when asked to review, audit, or improve existing Directive modules, systems, or orchestrators.

Generate Directive module scaffolds with schema, init, derivations, constraints, resolvers, and matching test files. Use when asked to scaffold, generate, or create a new module from scratch, or when the user describes a feature and wants the boilerplate created.

Test AI orchestrators without real LLM API calls using createMockRunner, write quality evaluations for agent output, add debug observability with tracing and metrics, integrate MCP tool servers, and wire RAG pipelines. Use when writing unit tests for agents, setting up CI evaluation suites, debugging orchestrator behavior, or connecting external tool sources.

Create AI orchestration systems using createAgentOrchestrator (single agent) and createMultiAgentOrchestrator (multi-agent). Define factsSchema with t.*() builders, init state, constraints for trigger conditions, and resolvers for LLM execution. Use when building LLM-powered workflows, agent pipelines, task runners, or multi-agent coordination systems.

Test Directive modules and systems using createTestSystem, mockResolver, assertFact, assertDerivation, assertRequirement, settleWithFakeTimers, and flushMicrotasks. Use when writing tests for Directive modules, constraints, resolvers, derivations, effects, or time-travel behavior.