@reactive-agents/scenarios

@reactive-agents/scenarios

Pre-built test scenarios and fixtures for Reactive Agents.

A curated set of agent test scenarios — each one is a task designed to provoke a specific failure mode (loop-prone, tool-failure, context-pressure, long-horizon, schema-drift) so you can evaluate whether the reactive harness keeps an agent on track. Every scenario ships its own task prompt, success criteria, expected failure-without-harness, and preferred models for reproducible benchmarking.

Install

bun add @reactive-agents/scenarios
# or: npm install @reactive-agents/scenarios

Usage

import { allScenarios, loopProneHaiku } from "@reactive-agents/scenarios";

// Run one scenario
const scenario = loopProneHaiku;
console.log(scenario.task);                    // the prompt to send the agent
const output = await runYourAgent(scenario.task);
const passed = scenario.successCriteria(output);

// Or iterate the whole suite
for (const s of allScenarios) {
  console.log(`${s.id} — expects "${s.expectedFailureWithoutRI}" without the harness`);
}

Each Scenario exposes:

interface Scenario {
  readonly id: string;
  readonly description: string;
  readonly task: string;
  readonly tags: readonly ScenarioTag[];
  readonly expectedFailureWithoutRI: FailureMode;
  readonly successCriteria: (output: string) => boolean;
  readonly preferredModels: readonly string[];
  readonly setup?: () => Promise<{ tools?: unknown; teardown?: () => Promise<void> }>;
}

API

• allScenarios — array of every bundled scenario.
• Individual scenarios: loopProneHaiku, toolFailureWebSearch, contextPressureNoisy, longHorizonRepoTriage, schemaDriftSql.
• Types: Scenario, ScenarioTag, FailureMode.

Part of Reactive Agents

This package is part of Reactive Agents — the TypeScript AI agent framework built on Effect-TS. See the full documentation.