code-testing-agent

# Code Testing Generation Skill Agent generates tests by matching project conventions, proving the tests compile, and proving the tests pass before completion. ## When to Use | Condition | Agent Action | |---|---| | User asks for new unit tests or broader coverage | Use this skill | | Existing tests only need execution | Use `run-tests` | | Coverage report analysis is the main task | Use `coverage-analysis` or `crap-score` | | MSTest-specific authoring or modernization is the main task | Use `writing-mstest-tests` | ## Strategy Matrix | Scope | Agent Strategy | Pass | |---|---|---| | Single file or small focused change | Direct generation path | Tests are written and verified without unnecessary sub-agents. | | Moderate multi-file feature area | One research → plan → implement cycle | Plan and tests cover the requested scope. | | Large project-wide coverage task | Iterative cycles over prioritized phases | Each phase finishes with passing verification before the next phase starts. | ## Pipeline | Step | Agent Action | Test | Pass | |---|---|---|---| | 1. Classify scope | Agent identifies language, framework, current test pattern, and request size. | Agent records source files, target test project, and existing framework. | Scope, framework, and target test location are explicit. | | 2. Research conventions | Agent reads source APIs, existing tests, and build or test commands. | Inspect selected source and test files in scope. | Conventions for naming, assertions, and test runner are clear. | | 3. Plan cases | Agent lists happy, edge, and failure cases per target file or feature. | Review generated plan or working notes in scope. | Every target behavior has at least one planned test case. | | 4. Implement tests | Agent writes test files that follow project structure and dependency patterns. | Inspect changed test files. | Tests match project conventions and target real behavior. | | 5. Verify build and execution | Agent runs the smallest existing build and test commands that cover the new tests. | Run existing build and test commands in scope. | Zero build errors and zero failing tests. | | 6. Run quality gate | Agent reviews remaining gaps and assertion quality by using related skills when the task scope benefits from them. | Run the chosen quality checks in scope. | New tests contain meaningful assertions and no obvious uncovered requested behavior remains. | ## Output Contract Agent delivers: - New or updated test files in the project test convention - Passing build and test evidence for changed scope - Short summary of covered behaviors and any explicit remaining gaps ## Verification Matrix | Test | Run | Pass | |---|---|---| | Build verification | Run existing build command for the touched project or solution. | Zero build errors. | | Test verification | Run the smallest existing test command that covers the new tests. | Zero failing tests. | | Coverage-shape verification | Inspect new tests or run existing focused quality tools in scope. | Happy, edge, and failure behavior each appear at least once in requested scope. | ## Guardrails - Agent writes tests that exercise existing behavior instead of rewriting production logic to fit the test. - Agent uses existing test frameworks and repository conventions. - Agent completes the task only after build and test verification succeed.