hive-test

What problem does it solve?

This Skill provides a structured framework to validate that AI agents meet their defined goals by running goal-based evaluation tests, surfacing failures, and guiding iterations.

Core Features & Use Cases

• Verifies that agents meet their goals through repeatable evaluation tests.
• Provides debugging support with the debug_test workflow to diagnose failures.
• Enables iterative improvements by re-running tests after modifications to agent code or goals.
• Integrates with hive-create and MCP tooling to generate test guidelines and orchestrate test runs.

Quick Start

To start, ensure your agent is exported to exports/your_agent and you have a goal_id and goal_json. Use the MCP tools to generate success test guidelines and then write and run tests:

• List existing tests with mcp__agent-builder__list_tests(goal_id=..., agent_path="exports/your_agent")
• Generate success criteria test guidelines with the MCP tool
• Write tests following the guidelines, then run tests with mcp__agent-builder__run_tests(goal_id=..., agent_path="exports/your_agent")
• If a test fails, diagnose with mcp__agent-builder__debug_test(goal_id=..., test_name="...", agent_path="exports/your_agent")