eval-agents

To test AGENTS.md completeness, spawn scenario-based evaluation subagents that grade agent responses against expected behaviors. This quantifies whether the documentation provides sufficient guardrails and context for reliable agent decisions.

Scenario-based evaluation for agent documentation is a testing framework that spawns subagents across scenario groups to grade responses and validate contextual guidance. It ensures documentation meets criteria for reliable agent decisions and health checks.

Evaluate AGENTS.md documentation after editing the file, adding or removing skills, or changing core workflows. This maintains alignment and ensures agents receive complete contextual guidance for their operational tasks.

Validate AGENTS.md completeness after changing core workflows by spawning evaluation agents across scenario groups to grade responses. This framework checks if updated documentation provides enough context for capable agent behavior and health checks.

Scenario-based agent testing works without external dependencies by using internal spawn logic to orchestrate subagents. The evaluation framework grades responses autonomously to quantify AGENTS.md contextual completeness and validate guidance criteria.