Test drizzle-solid features with the right level of proof. Use this skill when deciding whether behavior belongs in unit tests, integration tests, example verification, issue regressions, or parity tracking.
Use this skill when the task is about proving behavior, not only implementing it.
Current testing policy lives in docs/guides/testing.md.
Use these only as historical or planning inputs, not as the source of current requirements:
ACTION-PLAN.mdTESTING_STRATEGY.mdCOMPLETE-REPORT.mdEXECUTION-SUMMARY.mddocs/TEST-STRATEGY.mddocs/TEST-DIMENSIONS.mddocs/TEST-DIMENSIONS-COMPLETE.mddocs/TEST-COMPLETENESS.mddocs/DRIZZLE-ORM-MAPPING.mddocs/TEST-ANALYSIS.mdFINAL-SUMMARY.mdChoose the main test layer based on responsibility:
Do not treat CRUD, discovery, notifications, TypeIndex, or SPARQL behavior as mock-only concerns.
If a public example changes, verification should change with it.
Prefer a regression test plus any required docs/example update.
If a failure depends on planner, optimizer, hydration, or an unexpected execution-mode shift, do not stop at a generic feature test.
Add targeted proof for the interaction point, typically combining:
If the product rule is “keep exact-target behavior exact”, assert either the exact path or the explicit failure path. Do not reintroduce a widened scan just to make the test pass.
When using this skill, produce:
If the expected behavior is itself unclear, prefer opening or linking a kind:decision issue before adding misleading coverage.