skill-tester

Automate Claude Code skill validation by coordinating isolated tester agents and a persistent evaluator to verify triggering, workflow, resources, and output against defined criteria. It runs rigorous tests automatically to ensure new or updated skills behave correctly.

Yes, isolated test execution runs each test in a fresh tester agent to prevent context leakage. This ensures that validation results for multi-agent interactions remain accurate and independent across different test scenarios.

Validation supports positive, negative, edge, and story-based test scenarios to cover a range of real-world interactions. This verifies triggering accuracy, workflow adherence, resource usage, and output quality comprehensively.

Yes, the validation process satisfies requirements for frontmatter-based discovery. The evaluator checks skill metadata and triggering logic to ensure the skill is correctly discovered and activated by the system.

Structured reporting produces detailed evaluation reports with actionable recommendations for improvement. The persistent evaluator assesses output quality and generates reports highlighting areas needing attention after testing completes.

The best way to test edge cases is using isolated tester agents that apply negative and edge scenario tests against defined validation criteria. This end-to-end validation assesses workflow adherence and resource usage for robust skill performance.