quality-test-gen

Parallel test generation decomposes a codebase into a CSV of tasks and spawns concurrent agents to automatically produce unit, integration, and e2e tests. It classifies files, detects test frameworks, and generates failing-first tests following existing project patterns.

RED-GREEN TDD at scale is applied by spawning parallel agents from a tasks.csv to generate failing-first tests across many source files simultaneously. This targets directories where verification.json or coverage-report.json indicate missing or partial coverage.

CSV wave execution works by decomposing source files into a tasks.csv and spawning independent agents concurrently with configurable concurrency limits. It supports resume functionality to automatically produce canonical artifacts like results.csv and discoveries.ndjson.

Yes, you can generate tests automatically when verification.json shows partial coverage. The process targets phase directories with coverage gaps, runs suite verification, and summarizes coverage deltas alongside any discovered bugs in a final test-gen-report.json.

Parallel CSV-driven test generation produces canonical artifacts for verification and downstream automation, including results.csv, discoveries.ndjson, context.md, and test-gen-report.json. These files aggregate findings and report coverage deltas.

No manual test framework configuration is needed before generating tests with spawn agents. The process automatically detects existing project test frameworks and follows current stylistic patterns to avoid stylistic drift when generating failing-first tests.