Sposable — Test your Planning Partner
https://sposable.com/labs/planning-partner-test/

Missing case: ask for the essential scope and constraints before scheduling. Conflict case: flag the €50 overrun and request an authorized decision; do not reinterpret the ceiling or mark booking complete.

Example log: case → expected behaviour → observed output → pass/revise → changed instruction → rerun result. A passing normal case should still yield a useful output with its material caveats. If a failure persists, mark the configuration not ready for reuse and keep the failing input as a regression test.