Test Generation

Analyze source files, generate test suites, run them, and iterate on failures until the tests pass.

Workflow

runPipeline(
  pipe(
    listSourceFiles,
    forEach(
      pipe(
        analyzeForTestability,
        generateTests,
        loop((recur) =>
          pipe(runTests, classifyResults).branch({
            Passing: drop,
            Failing: pipe(fixTests, recur),
          })
        ),
      )
    ),
  ),
);

Stages

1. List source files — find files that need test coverage.
2. For each file (concurrently):
   • Analyze — identify functions, edge cases, and testable behaviors.
   • Generate tests — write a test file based on the analysis.
   • Run and iterate — run the tests. If any fail, fix them and re-run. Loop until all pass.

Key points

• The analysis agent focuses purely on identifying what to test — it doesn't write tests.
• The test-writing agent receives the analysis output, not the analysis instructions. Focused context.
• The fix-and-retry loop is the same loop + branch pattern used in type-check/fix.
• forEach processes all files concurrently, so test generation scales with the codebase.
• Consider adding withTimeout to cap how long the fix loop runs per file.