Test generator (three subagents)

When to use

• Missing tests or low coverage
• Before refactoring (safety net)
• After implementing a feature
• Explicit requests to add or generate tests

Goal

Generate realistic, stack-aligned tests and run them only after explicit user approval.

Pipeline overview

Parent Agent
  -> Subagent 1: Stack Analyzer (write /tmp/test-context.md)
  -> Subagent 2: Test Generator (write tests, /tmp/test-manifest.md)
  -> Pause for user approval
  -> Subagent 3: Test Runner (write TEST_REPORT.md)

Phase 1: Parent reconnaissance

1. Identify scope from the prompt (file, module, or test type)
2. Inspect project config (package.json, pyproject.toml, etc.)
3. Find existing tests and conventions
4. Provide scope/context to Subagent 1

Phase 2: Subagent 1 (Stack Analyzer)

Detect language, test frameworks, conventions, and target files. Write /tmp/test-context.md.

Phase 3: Subagent 2 (Test Generator)

Use the patterns in references/test-patterns.md. Write tests and record a manifest in /tmp/test-manifest.md with file paths, test types, coverage, and run commands.

Phase 4: Pause and confirm

Summarize generated tests and ask whether to run all tests, only a category, or none.

Phase 5: Subagent 3 (Test Runner)

Run approved suites, capture failures, and write TEST_REPORT.md. Clean up temp files.

Notes

• Use realistic domain data in tests.
• Align with existing test style when present.
• Never run tests without explicit user confirmation.