Repro
$ ./mxcli test tests/ -p eShopMendix.mpr --require-assertions
Error: unknown flag: --require-assertions
$ ./mxcli test --help | grep -c require-assertions
0
But .ai-context/skills/test-microflows.md — written by mxcli init from the skills embedded in
this same binary — documents it:
# Fail the run on any test that asserts nothing
mxcli test tests/ -p app.mpr --local --require-assertions
Why it matters
The skills are embedded in the binary and rewritten by mxcli init --sync-skills specifically so
guidance and binary stay in step — the bootstrap script even comments that upgrading the binary
"used to leave yesterday's guidance in place with no warning — and an agent reads stale guidance
with the same confidence as current guidance." Here they are out of step in the same release, which
is the failure mode that machinery exists to prevent.
Either the flag regressed out of the test command, or the skill documents an unreleased feature.
A vacuous-test guard is genuinely useful, so landing the flag seems better than deleting the doc.
Repro
But
.ai-context/skills/test-microflows.md— written bymxcli initfrom the skills embedded inthis same binary — documents it:
Why it matters
The skills are embedded in the binary and rewritten by
mxcli init --sync-skillsspecifically soguidance and binary stay in step — the bootstrap script even comments that upgrading the binary
"used to leave yesterday's guidance in place with no warning — and an agent reads stale guidance
with the same confidence as current guidance." Here they are out of step in the same release, which
is the failure mode that machinery exists to prevent.
Either the flag regressed out of the
testcommand, or the skill documents an unreleased feature.A vacuous-test guard is genuinely useful, so landing the flag seems better than deleting the doc.