Pass Through Wall joins the march: the bot banks a crossing when stepping
through one wall beats the walk by 4+ steps (or no road exists), walks to
the chosen wall, and spends the charge stepping through. passWallCharges
is now in PlayerPublicView — the cast is public at a physical table.
Path denial for guardGold tiers: when an enemy carries the bot's treasure
toward home, or closes on its gold on the floor, the bot reconstructs the
threat's shortest path and prices every corridor line and empty square on
it; CREATE WALL, FILL SQUARE WITH STONE, or THORNBUSH lands wherever the
detour costs the enemy 3+ steps, legality mirrored from the engine
(sighted, empty, off homes and warp tokens).
All five cards leave the bottom discard tier (2 -> 6).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0138A8CjeQRpvzKxuMfz1Bqc
Rules rev 24: a waterwall wave that finds a victim dist cells from its
source has only range-dist spaces of push left — a range-2 wave throws its
adjacent victim two spaces but a victim at its far edge only one, and
spent force never converts into crush damage. Applies to WATERWALL and
both STONE TO WATER waves; older revisions keep the flat full-range wash
so stored games replay unchanged.
Automatons now use DESTROY WALL on the march: two BFS distance maps (from
the bot, from its objectives) price every visible wall by the shortcut its
removal opens; the bot blasts when it saves 4+ steps of walking — or when
no road exists at all — never standing beside the blast unless trapped
and healthy. The card also leaves the shed pile: discardValue 2 -> 6.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0138A8CjeQRpvzKxuMfz1Bqc
An automaton answered NO SPELL with an Absorb — which "has no effect
on duration-based spells." Afflictions sat outside the price table
and drew the phantom two-point default, tripping the point-soaking
rungs. An affliction's weight is now its duration, and the two
counters that touch only points — ABSORB and REVERSE — stand aside
for it, leaving the ones that genuinely bite durations: full shield,
remove curse, blunt, reflection, shieldstone numbers. Pinned: under
NO SPELL with both in hand, the bot blunts, never absorbs.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Scoped to everything since the last pass (3308850). Three blind
reviews, every finding verified before touching anything.
Confirmed and fixed: two identical comment-splitting insertions left
doc comments orphaned from their fields (net and local alike); a
51-line CSS fossil of the pre-extraction inline effects survived in
Board.svelte; the rev-13 miss-roll test asserted tautologies while
its comment claimed a check the code never made — it now proves the
die was consumed, and the skeleton is no longer returned as trollId;
the sprite registry's `as never` silently disabled the completeness
its annotation advertised (now a mapped type, one cast at the
dispatch seam); a dead ternary guarded a union that doesn't exist;
Bolt carried a duplicate .fork rule from a color iteration; fxTtl
contradicted three sprites' real animation lengths; the permanence
sentinel was reinvented as a magic 9000 (the engine now exports
isPermanentDuration); CELL was declared thrice (fx.ts now imports
it); App and Replay ran two divergent fx schedulers (one scheduleFx
now, cancellable — stale flourishes can no longer fire after leaving
a game); TokenArt retried missing files forever; the anti-anti
escape guards merge with the gate asymmetry explained; wall-of-fire's
rev-12 carve-out is marked; overLimit ignored a displayed BRAINSTONE
(bots over-discarded by two); botRemark's header mis-stated its own
branches; deliverGold fired on any drop, not a home-base delivery;
escape and win banter never fired from the steps that carry them.
Rejected: "as a human would" (house voice); FxGallery's dev-harness
framing (trimmed one plea, kept the facts).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Two archmages idled turn after turn: hands clogged with situational
neutrals, and doEndTurn caps the draw by the room left, so a full
hand never refreshed. Nothing to cast, nothing to draw, nothing done.
The brain now sheds dead weight before ending the turn — only cards
it has no play for; good counters and attacks stay hoarded — so the
draw always has room. The trick is repertoire, so it is tier-gated:
adepts and archmages churn, the apprentice hoards like the novice it
is. Measured across forty berserker mirrors that sharpened the
handicap from two-to-one to three-to-one with zero stalled games.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
An automaton answered DROP OBJECT with an Absorb, then a Blunt —
two counters drained against a spell that carries no points. The
brain's price table listed only the classic damage spells; anything
unknown was guessed at 2 points, which tripped every damage-counter
rung. Utility attacks (dropping, dragging, swapping, peeking) now
route past those rungs entirely: the bot passes, or spends a full
stop when the effect actually threatens something — a shield for the
treasure it carries, a worrier's refusal to be dragged. Stone Dead,
Butt Head, Heave-Ho, and Illusionary Attack join the price table
while the hood is up.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Scoped to everything since the last pass (695307d). The residue of
fast iteration, removed: a reduced-motion media query that had
swallowed a full copy of the .faq-seal rules; doc comments orphaned
from their functions by inserted methods; the FAQ scrape's seams
(section headings run into ruling bodies under the wrong topics, a
next-page heading shipped as a ruling, an amputated "h", and rulings
filed under alphabetically-nearest strangers — Large Rock/Dagger now
lives on those cards; Book of Spells and Torquemada describe no card
in this set and are gone); the write-only aisle flag left behind by
the reverted rev 7; a linter-silenced dead destructure; a duplicated
median lookup; dead casts; and the fallback path that ignored the
apprentice's one-card draw.
Tests now typecheck (tsconfig includes test/), which surfaced the
missing type imports and a drifted creature literal hiding under an
as-cast. Also: a no-op self-assignment, mid-file imports hoisted, a
dynamic-import habit made static, a hedge comment replaced with a
loud setup failure, the fallback-retries-itself dead rung removed
from both bot drivers, and the twin peek scrims merged.
Deliberately kept: the TIERS lookup guard (ledger JSON is untrusted),
the wand-cost stanzas (they differ on the power-attack trade — a
rules question, not a dedup), and rollDie's coexistence with rollD4
(migrating changes visible logs; future work).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Tiers degrade resources and repertoire, never judgment — the design
constraint was that no tier may ever look dumb. The APPRENTICE draws
one card a turn instead of two (a poorer wizard, not a worse one),
spends counters only on heavy hits (thrift, not blindness), and
carries a modest spellbook: damage, summons, stones, keys, and
treasure play, with no afflictions, amplifies, ambushes, guarding
tricks, or deja-vu. The ADEPT draws fully and knows everything except
ambushes and amplify. The ARCHMAGE is the full curriculum. Every card
any tier plays, it plays correctly.
The lobby workshop gains a tier picker beside the temperaments
(default adept); the tier persists with the seat, shows in roster and
scoresheet ("⚙ apprentice mystery"), and rides the drive loop.
Measured where it should matter: in berserker combat mirrors the
archmage beats the apprentice two to one, while pure treasure races
stay honest — the handicap lives in the card exchanges a human
actually feels, pinned deterministically in the tournament suite.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The automaton now plays number cards where they matter: attached to
value-scaled attacks (waterbolt, powerthrust) choosing the biggest
for damage, and played for movement when the goal is just out of
stride. It raises its monsters when no wizard is in spell range and
commands them every turn — creatures march at enemies by the same
hazard-shy BFS and maul whoever shares their square. Targets are
picked by lowest life.
And it has moods. The HUNTER plays the classic game: gold first,
violence when convenient. The BERSERKER hunts wizards over treasure
and finishes games by last-wizard-standing. The WORRIER paths around
enemies, counters at a lower threshold, shields against chaos, and
never brawls. Hosts pick the temperament when seating one (or the
workshop assigns at random); it persists with the seat, shows in the
roster, and the tournament harness proves every pairing finishes.
Across ten seeds of berserker versus hunter: six treasure wins to
four kills — the moods play genuinely different games.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Phase 4 begins. The automaton is a pure function in the engine —
automatonCommand(view) — playing from its own redacted GameView, the
same information a human seat receives: hidden hands stay hidden from
the clockwork. It ranks simple damage spells, counters what hurts
(full shield at 3+, reflection at 4+, blunt at 2+), discards its
worst cards by a value order, BFS-pathfinds to enemy treasures and
home again, refuses to path through hazards, brawls when there is
nothing to steal, and always has a safe fallback; the server's drive
loop steps any bot-held seat through the same runCommand path as
humans, so bot commands log, persist, replay, and broadcast like
anyone's.
Hosts seat them pre-start with "⚙ seat an automaton" (Automaton,
Automaton II, ... V); bot seats persist as tokenless join lines and
restore on boot. Proven three ways: bot-vs-bot engine games conclude
across seeds in 8-14 rounds (~100 commands — human-scale), a full
four-automaton table finishes, and a live websocket game of human
vs. automaton ended with the clockwork carrying two treasures home
through a do-nothing opponent.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Eric WagonerandClaude Fable 5