Microsoft has open sourced code-testing-generator, a polyglot unit-test agent shipped within the MIT-licensed dotnet/skills repository. The tool autonomously plans, writes, runs, and verifies unit tests across multiple languages and frameworks. The agent addresses vague “generate tests” prompts by scanning repositories, inferring languages, frameworks, and conventions, and locating real build and test commands. It follows a Research-Plan-Implement pipeline, avoids touching production code, and blocks flaky or environment-dependent tests. Internal benchmarks show 92.1% task completion versus 78.9% for stock GitHub Copilot using the same model, especially on vague and diff-focused prompts. The project targets startups, regulated industries, and enterprises seeking higher reliability and standardized testing workflows.
This update represents a notable development in the Ai sector. Organizations and founders tracking this space should evaluate potential strategic and technical implications on their operations.