The Agent Skills Directory

Automated Build and Test Execution: The skill coordinates several sub-agents to compile code and run tests during the generation phase. This is an intended part of its functionality to ensure that the produced tests are workable and follow project conventions.
Evidence: The implementation phase in SKILL.md describes using code-testing-builder and code-testing-tester to verify that the generated code compiles and passes.
State Management: The skill utilizes a local .testagent/ directory to store research results and implementation plans. This allows for persistent state management during complex multi-agent tasks.
Evidence: SKILL.md identifies files like research.md and plan.md stored within the .testagent/ folder.
Input Analysis Surface: To generate effective tests, the skill must read and analyze project source files. While this creates an interface for processing external data, the skill includes explicit instructions to use mocks for all external dependencies (HTTP, database, etc.) and to focus on unit testing.
Evidence: The unit-test-generation.prompt.md file contains strict requirements for mocking external services and avoiding tests that bind to network ports or call external URLs.

code-testing-agent