In the wall-framing takeoff CLI, --panel-max=zzz did not fail. The size parsed to zero, the panel feature switched itself off, and the tool printed an ordinary order. A framer who mistyped 12ft would get a normal-looking result and no panelised walls. --rim-lap and --sheathing-psf did the same.
All three now stop with a message that names the flag, shows the bad value and says what is expected, such as a length such as 4ft or 9.25in. A new test covers five bad values and checks that good ones such as 12ft, 2.5 and 9.25in still run. The 34 tests in cli.test.mjs and cli-fuzz.test.mjs pass.
The fix is small. Quietly accepting bad input in a quantity tool costs more, because the wrong output looks right.