improving-test-suites
Improving Test Suites
Portable orchestrator that turns a named test suite into the smallest useful behavior-focused harness. Tests are executable contracts: a test earns its place when it would fail for a real break in public behavior, validation, security, meaningful failure handling, or a production-relevant edge case.
The orchestrator serves confidence and safety, not test count. It delegates raw work to subagents, keeps compact reports, gates destructive changes, verifies approved behavior coverage survived, and returns exactly one handoff status.
Portable target: OpenCode and Claude Code. Plain Markdown links and minimal frontmatter only. If the runtime cannot spawn subagents, execute each named definition inline as a strictly scoped pass, then retain only its report.
Inputs
| Input | Required | Example |
|---|---|---|
TARGET_TEST_FILES |
Yes | tests/test_billing.py, tests/api/ |
USER_GOAL |
No | reduce brittle implementation-coupled tests |
TEST_COMMAND |
No | pytest tests/test_billing.py -q |
SCOPE_LIMITS |
No | test files only |
REFERENCE_NEED |
No | pytest parametrization |
AUTO_APPROVE |
No, default false |
true only for explicit headless plan-gate bypass |
RESUME_PACKET |
Conditional | Packet from COMPLETE_BLOCKED |