Trap 12 told only the first act. The full story is the lesson: ThemeKeys.ALL excludes
the font roles by its own docstring, and the builder styles some base types directly
(RichTextLabel — which renders every line of DM prose on the creation screen). The
meta-guard written specifically to prevent a third miss contained a `continue` that
exempted exactly the gap it existed to catch.
A guard-of-a-guard with an exemption in it is not a guard.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The creation screen is done — 302 client tests (up from the 250 baseline),
content build green. Also flips the M3 Title screen from planned to done: it
shipped in a77bf03/f071392 and the roadmap never caught up, which would have
misled M4-c into thinking it must build one.
Adds six new trap species to docs/traps.md from this milestone's review
passes: substring-collision assertions, a BBCode wrapper masking an empty
ContentDB, a test asserting the script's own loop bound, a trivially-true
gating fixture, assertions already true before the action under test ran, an
"iterates everything" guard that checked three hardcoded pairs, a node-type
sweep that missed RichTextLabel, tests that called handlers instead of
pressing nodes, GUT silently skipping an unparseable class_name test file at
the old count, and a test that can hang instead of fail. Also documents the
new client-docs convention: clickable cards are Buttons with mouse_filter = 2
on every child, and pressed on a toggle_mode Button is the chosen state.
Human F6 confirmation of the rendered screen against the mock is still
outstanding and is not claimed here.
Six bugs this project shipped into a branch, what let each through, and the
guard that now catches it. They share a shape: the test passed, and the test
was wrong.
- a regression test that constructed its own correct object, so reverting the
bug left the suite green
- GDScript default params turning an arity change into a silent miscompile,
invisible to a grep for the renamed field (grep cannot see positional args)
- a determinism test comparing two runs that would change together
- a closed move vocabulary that was not bounded, because the model could repeat
a tag and choose an amount in unary
- a roster written down in four places with only one pair guarded — including a
shipped authoring skill that would have emitted origins allowing zero callings
- generated content that has a client consumer
Linked from CLAUDE.md §18 and docs/README.md. Also records M4-a on the roadmap
with its live end-to-end proof.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QYa9u7Kdxv5gX4AnwWexy8
