Let the AI use Venser, the Sojourner (and two other mass-unblockable cards)

[liamiak]

28 added lines outside the test, which is a droppable second commit.

Venser, the Sojourner is AI:RemoveDeck:All. Two of his three abilities turn out to work already —
the flag was covering for one dead one, and that one turned out not to be his alone.

What was dead

The -1 never fired on any board, including one where it wins on the spot:

AI power	Opponent life	Blockers	Before	After
18	15	2	CantPlayAi	activates
18	18	2	CantPlayAi	activates
18	19	2	CantPlayAi	declines
6	20	2	CantPlayAi	declines
no creatures	20	2	CantPlayAi	declines
18, but Main 2	15	2	CantPlayAi	declines

While there are corner cases there unblockability without being able to swing for the win is worth it, those are relatively rare and not implemented. However, the case of "can I make everything unblockable and swing for the win" is key, and now implemented.

UnblockableForLethal holds the ability until attacking with everything that can attack would take
an opponent to zero with nothing able to block — the state the ability itself creates, so blockers
are not counted.

The same gap on two other cards

Every card where granting mass unblockability is an optional choice had the identical problem —
CantPlayAi on every board, lethal ones included — and all three now fire only on lethal:

Card	Ability	Flag
Venser, the Sojourner	-1	AI:RemoveDeck:All, removed here
Jace, Arcane Strategist	-7 ultimate	AI:RemoveDeck:Random, left alone — that is a deck-building signal, not a playability one
Tezzeret's Gatebreaker	{5}{U}, {T}, Sacrifice	none — this one was simply misplaying in AI decks

Those two argue for the gate harder than Venser does: Jace spends seven loyalty and Gatebreaker
sacrifices itself, so firing either off-lethal is a real loss rather than a wasted tick.

The rest of the cards with the same effect are deliberately untouched, because using it is not a
choice they make: Keeper of Keys and Veiling Oddity are mandatory triggers, and Bedlam, Tazeem and
Tanglewalker are static. Glaring Spotlight grants hexproof alongside it, so it is also a protection
card and a lethal-only gate would be wrong for it. Reverse the Polarity offers it as a Charm mode,
which CharmAi picks through a different path and would want its own verification.

No priority plumbing was needed — ComputerUtilAbility already sorts planeswalker abilities
most-costly-first, so the -1 outranks the +2 by itself once it starts saying yes.

What already works

The +2 (exile a permanent you own, return it at end of turn) lands in ChangeZoneAi's existing
blink logic, because the ability has Destination$ Exile with a DB$ DelayedTrigger sub. That logic
is good: it skips tokens, prefers permanents with ETB triggers, holds until post-combat, and takes
anything the AI owns but an opponent controls immediately rather than waiting. Measured:

Board	Phase	Result
Opponent controls my Wall of Omens	Main 1	takes it back
Opponent controls my Grizzly Bears (no ETB)	Main 1	takes it back
My Wall of Omens	Main 1	declines, waits
My Wall of Omens	Main 2	blinks it
My Mulldrifter	Main 2	blinks it
Creatures with no ETB, or lands only	either	declines

The -8 is already AILogic$ Always, and the emblem it makes behaves: whenever the AI ultimates,
the emblem exiles Ancient Brontodon and then Grizzly Bears — the opponent's board, biggest first —
leaving its own Colossal Dreadmaw alone.

Testing

• mvn -pl forge-gui-desktop -am test: 287 tests, 0 failures.
• The test fails without the change, in both directions.
• It asks the AI directly rather than playing a game out, so it does not depend on MyRandom.

The flag comes off on the back of that.

🤖 Implemented with the assistance of Claude Code (Opus 5).