Hold the Flood — Combat Rules

CMB101 / CMB102 · r14 · 2026-09-07 · Combat and Enemy Designer · In review. r14 answers the two owner corrections that land on this page's foundations, in a new §0C. "Water is not a requirement": four of seven roles and the coil are water-defined, so [R14] proposes stating the shared rules over a conductive medium, with water as the pumping district's instance — no behaviour and no role changes. "A one-off authored room wipe does not by itself demonstrate build expression": this retires the evidence r12 used to hold R8, and the page says so rather than waiting to be told; [R15] proposes in its place that build expression scale through targeting geometry — hops, reach, origins, eligibility, re-entry — which R8 does not forbid and which is portable by construction. §12 was revised by the Visionary while r13 was current; the disposition is right and this revision does not contest it — only the general "no armoured class" statement is moved into §6, where the rest of the chain geometry lives. Sources read: Design Document r14 ec1b6c §§3–6, 8, 12–16 and Appendix A; the shared-rule contract Power and Attraction r2 69024c and its r1 seed 8afc3c; Campaign 34dc09; Design Document r15 5f9fbb; Design Review REV104 r1 80a882; Enemies c63228; Cutter; Equipment; World and Mystery r20 7699de; Expansion Brief; Design Index dd01c7; Tasks eb8c03; PMR003; Design Review REV103 r5; Brief D013/D016/D019–D021; Reference/Content Scale REF107; External Critique EXT102; Wiki Conventions. Design Index · Expansion Brief · Glossary · Enemies

Nothing on this page has been built, prototyped, simulated or played. Every number is a proposal. Where I check a written scene against the rules, I am checking arithmetic and precedence on paper, not reporting behaviour.

What this page is for

The Lead owns the spine rule — conduction, power, attraction, charge — and as of Power and Attraction r2 69024c that page is the shared-rule contract, with Design Document §4 as its derivation. That rule is canonical there, not here. This page does three things the spine rule deliberately does not:

1. States the order in which a body resolves competing reasons to move, so that a roster of more than one enemy is designable at all.
2. States the shared grammar of targeting, damage, status and charge, so that Equipment and this page cannot invent two incompatible systems.
3. Proposes ten numbered revisions to gaps the roster exposes, each with a readable signal, a cost, and the choice it improves.

Where I restate an adopted rule it is a copy for legibility and the Design Document wins. Where I propose something it is marked [Proposal] and numbered.

0. Current state: what is adopted, proposed, or withdrawn

This section replaces the two stacked reconciliation appendices r2 and r3 added. PMR005 asked for one current contract rather than another layer, and it is right: the older narratives are in page history and are summarised in the revision log at the foot of this page.

Where the rules live. The Lead owns the spine rule, and Power and Attraction r2 69024c is the shared-rule contract — the page to cite for conduction, power, attraction and charge. Equipment owns items. This page owns the order in which a body resolves its reasons to move, the targeting/damage/status grammar, and composition. Where the three disagree, the contract wins on rules and Equipment wins on its own items; this page names the disagreement rather than routing around it.

0.1 The intensity rule

Everything PMR004 asked to be separated — fixed power, the coil, the arc, a Spark's death — separates on one principle the design already has. GDD §4's grounding note observes that real electrofishing uses one electrode at two intensities, a weak field that draws and a stronger one that stuns. Applied to bodies:

Current does two different things to two different kinds of body. To the surveyor, energised water is damage — one health per second, prevented by insulation, at any intensity. To an organism, ordinary energised water does nothing: an organism is made of charge, and the water is weaker than it is. Only a bank-strength discharge reaches an organism, and what it does is stun and make it shed, never damage. Nothing in this game damages an organism with electricity. The arc damages by conducting the cutter's hits along bodies — a weapon effect with a weapon's cue, not a water effect.

This is a proposal to the Lead, not a deduction: it is the smallest rule that answers all four of PM's questions at once, and it is what stops a Spark's death quietly supplying a chain-kill rule.

0.2 The effect contract — Combat's half, for Equipment to countersign

PMR004 asked Combat and Equipment for one table. This is Combat's half, written to match Coil C1 and Arc A1 rather than against them. Where Equipment had a number and this page had a different one, Equipment's is adopted and the row says so. It is not agreed until Equipment says so.

Boundaries between overlapping pulses — proposed by Equipment, adopted here as Combat's rule. A paid discharge samples the conducting bodies in its wet footprint at the instant it fires; bodies that arrive later do not inherit that stun from the two seconds of ordinary power. A fresh paid pulse resets an eligible survivor's recovery to eight seconds from that pulse — it never adds durations and never resets spent state. Removing a connection stops power crossing it but does not cancel a stun already applied. A temporary source expiring removes only its own contribution: fixed power, or a later Spark pulse, can keep the water hot. An arc burst is legal and useful in a channel with no power at all, because it never touches the water.

Fixed power and the branch's water — ruled, and this page was on the losing side. Contract §2: a lit branch's channel water is energised for as long as the branch is lit, "which is why the electrician waits in the lit yard in insulation". This page recommended the opposite and the recommendation is withdrawn. The objection was that the two-second ruling's rationale — collecting without compulsory insulation — fails in a lit branch; it does not, because that rationale is about water no lit branch feeds, which is the Tidal Barrier's flats and the Lock Flight's head basin rather than the electrician's yard. So in the pumping district's lit yard, insulation is the entry fee and the coil's two seconds sit inside a hazard already running (Encounter A).

0.3 The roster, and who owns which part

Eight roles. Three are the Lead's, on Campaign; four are this page's; one is a merge. PMR005 accepts the Walker naming for coordination and treats the merged roster and the bounded Carrier as proposals, which is what they are.

The roster is eight, and the contract's §7 now lists all eight as Combat's, with the Floater transferred here. r5 stops this page saying "six" anywhere.

Corrected in r4: r2 of this page said all four Campaign roles sat at line 2. They do not, and PMR005 caught it — only the Floater does, and only while its water flows. Ordinary Walker, Heavy, Crust and Spark attraction is line 3; a role sits at line 2 only when its target is something other than a beacon.

0.5 The Spark row — ruled, and a crossing this page has to own

Settled: a Spark killed in water energises that water for two seconds, stuns nothing, damages no organism, makes no chain, and hurts an uninsulated surveyor standing in it. Its own kill mote is the yield if it had not already shed. That is contract r4 §4, and the Lead chose it as this page wrote it.

How this page then got it backwards, which is worth recording rather than quietly fixing. The Lead ruled at 19:17. At 19:23 r5 of this page, reading the contract at r3, adopted the old row "rather than silently contradicting the contract" — so the contract held Combat's rule and Combat held the contract's, each following the other into the position the other had just left. REV104-15 caught it. The lesson is narrow and useful: when mirroring another page's rule, re-read that page at save time, not at draft time — the same guard this wiki already applies to text, applied to a rule.

The argument that won, kept because it is the reason and not the ruling. A Spark that stuns and sheds is a free coil: the coil is two cells plus a full ten-unit bank, and stopping a wet crowd and making it shed is exactly what that buys. A free version would strictly dominate it wherever Campaign places Sparks — and dominate it by an event the player does not choose, because the cutter cannot be withheld (R6). It would also spend the once-per-organism yield without a decision, and that rule is what the charge economy rests on. Equipment recommended the same independently.

0.6 Presentation categories — deferring to Campaign §7, with one difference

Both this page and Campaign §7 published "the one list" of presentation categories inside the same quarter-hour, which REV104 r5 noted without comparing them. The Lead's §7 says it is the one, REF110's audit is already running against it, and this page defers to it. No competing list is kept here.

They agree on the thing that matters, which is the answer to PMR006's question: the roster's cost is not eight creatures. It is one base body plus one licensed edit, with everything else a pose, a scale or an engine-drawn cue.

One difference, and the Lead's reading is the better one. This page counted the Spark as a second licensed edit; Campaign §7 puts its flicker under engine-drawn cues, leaving the Crust's encrustation as the only edit. Campaign is right — a flicker is motion on the base body, not a changed sprite, which is what this page's own R4 says when it insists the Spark reads by motion rather than by brightness. So the roster costs one body, one edit, five poses-or-scales, and the rest engine-drawn.

Combat's contribution to that audit, which stands: which body each cue must be readable against — a crowd of forty to a hundred and forty, at a glance, with four facts competing for cues (§0.4, C9, now ruled on the contract). Nothing has been located, inspected, licensed, downloaded, purchased or created by this role.

0.4 What is still open, stated as questions with my position

0C. Two owner corrections, and what they cost this page

Both land on this page's foundations, and neither is answered by renaming anything.

1. "The Flood means massed enemies; water is not a requirement derived from the title"

The exposure, counted honestly. This page's spine is conduction, and conduction here runs through water. Of the seven roles, the Anchor (lowest wet ground), Spark (energises the water it dies in), Crust (does not conduct) and Floater (rides flow) are water-defined, and so is the coil, which is the paid action everything else is priced against. If water is not a requirement, four of seven roles and the design's signature verb are sitting on an assumption the owner has just declined to make.

The repair is one word, and it costs nothing. Every rule on this page is stated over water when what it actually needs is a conductive medium that connects bodies. Water is one instance of that and a good one; it is not the only one. Flooded ground, a rain-soaked yard, a spill across a plant floor, wet slurry, a sheet-metal deck, a coolant flood — each supports the identical grammar: a connected footprint, a paid pulse that energises it, bodies in it that conduct or do not, and a surveyor standing in it who pays for being there.

Proposed [R14]: state the shared rules over "conductive medium", with water as the pumping district's instance. Nothing else changes — not the precedence stack, not the effect contract, not one role's behaviour. The Floater is the only role that needs flow specifically, and flow is a property a medium may or may not have, so it becomes a role a district can field or not, exactly as the composition budget already treats it. This is the Lead's to rule — the medium is the shared contract's word, not mine — and I am proposing it rather than editing around it.

What I am not doing: not renaming water to sound less watery. The owner's correction says the requirement is wrong, not the word. A district that wants a flooded yard should still have one.

2. "A one-off authored room wipe does not by itself demonstrate build expression"

This retires the evidence I was holding R8 with, and I should say so before someone else does. Last revision I argued that no damage ladder was needed because the Choir Cascade reached a spectacular clear with no damage value anywhere in it. Equipment has since moved the Cascade into the place — no portable item, a tuned furnace-hall network, a finite authored crowd — which does deliver the wipe, and which is precisely what PM's correction says does not demonstrate build expression. So my evidence for R8 no longer exists. The ruling it was built on survives and was adopted: the network strikes each Coil-stunned ordinary body inside its footprint once and kills it is the kill-a-stunned-ordinary-body rule, and Equipment's bounding of the crowd instead of the effect is a legitimate answer to the tradeoff I named. It is just not a build.

So the honest question is now open: where does outrageous portable power come from, if not from damage numbers? The owner wants "portable component interactions that turn a cramped rack into outrageous power". This page has exactly one general multiplier — the arc chain — and it is bounded by hops, reach, targets per hit, and eligibility. Those are targeting quantities, and targeting is this page's to own.

Proposed [R15]: build expression scales through targeting geometry, not through damage. R8 forbids damage values, resistances and multipliers. It does not forbid a component from changing where a chain can go, and that is a large, unexplored space that costs no new concept:

Stacked, those are outrageous, and none of them is a bigger number on a weapon. A rack that turns three hops into five, doubles reach and seeds a second origin has not increased damage at all — it has changed the shape of what one automatic hit does, which is the "ridiculous build works" reaction the owner described, and it is portable by construction because it lives in the rack.

This is a proposal to Equipment, whose catalogue it is, and to the Lead. I am not adding items. And the honest limit: if targeting geometry turns out not to reach "outrageous", then R8 is the thing to reopen, and it should be reopened deliberately rather than eroded by one exception at a time. I no longer have evidence that R8 is free, and I would rather record that than keep asserting it.

1. The central risk, answered first

The strongest thing about this design is that one rule answers four questions, and every reviewer on this wiki has said so. A roster of enemies is the most likely way to wreck it. The usual way a survivorlike grows a bestiary is by adding rules — this one flies, this one shields, this one summons — and after six of those the player no longer has one rule, they have a lookup table, and the district stops being readable.

So the roster is built on a constraint rather than a licence:

A role may only subtract from the spine rule, or restate an existing clause about itself. It may not add a new clause.

Concretely, the six roles in Enemies introduce exactly two new facts about bodies, and both are already in the player's vocabulary:

Everything else a role does is not answering the light: it targets something else, or it stops something else being lit. The Flood's mass — 85% or more of every surge, by the composition budget in §7 — obeys the attraction clause exactly as written. A player who knows only the spine rule still predicts almost every body on screen, and the exceptions announce themselves before they act.

The credible alternative is the one the genre normally takes: give enemies stats, and give the player numbers to beat them with. I am against it for a reason specific to this game rather than for taste. This design has no damage curve at all — no XP (GDD §6), a fixed cutter, and an arc that is a spend bounded three ways. A stat ladder would have to be invented from nothing to give stat enemies something to be beaten by, and it would immediately compete with the thing the design actually sells, which is that understanding changes the outcome and nothing else does. See R8.

2. The precedence stack [Proposal R1 — a clarification, not a change]

GDD §4 states pursuit range and attraction in one paragraph and does not order them: "Two things move organisms besides beacons: pursuit range… and press." With one enemy type that is enough. With six it is not — and neither is it enough for the existing scene in §13, where the player stands inside a crowd that is also being pulled by a beacon.

Every body, every frame, resolves the first line that applies to it:

Two consequences worth stating, because they are the ones a designer will get wrong:

• Pursuit outranks attraction. A crowd holding at a beacon still peels off to chase a surveyor who walks into it — which is what the cutter fires at, and why harvesting motes inside a held crowd costs health. GDD §4 already implies this when it says harvesting means "working at the crowd's edge inside its pursuit range". R1 only writes the order down.
• Line 2 sits below pursuit and above attraction. A Carrier standing next to you chases you as a body, not as a battery, and at that distance the two are indistinguishable, so nothing new has to be read in a scrum. The role only shows at range, which is where the player has time to read it.

Cost of R1: none to the player; it removes an ambiguity. Risk, now bounded: if pursuit range is generous, line 1 swallows lines 2–4 near the player and the roles read as identical up close. Contract r4 §3 gives pursuit range a value — about ten metres — which settles the shape of the risk rather than the risk itself: ten metres is short enough that a role's own target still governs most of its walk, and the Carrier's twelve-metre acquisition sits just outside it, so a Carrier reads as a role before it reads as a pursuer. Whether ten metres feels short is unverified and unverifiable here.

Line 4, generalised [Proposal R12, new in r7]. The contract and GDD say a crowd with no reachable lit beacon spreads along the branch it entered by. That wording is exact for the pumping district, where a crowd never leaves its branch. It stops being exact the moment a portable beacon pulls a crowd somewhere else and then dies — which is precisely EQP102's third approach, where a lure is planted on a side apron and expires there. Read literally, those bodies would have to spread along a branch they are no longer on.

Proposed: a crowd with no reachable lit beacon spreads outward from where it is, along the ground it occupies. For a crowd that never left its entry branch this is the same sentence. For one that was moved, it is the only reading that is not a teleport. It adds nothing a player must learn — "no light, so they spread from there" is what anyone would predict — and it makes the planted-lure play legal instead of undefined. Lead's to rule; Equipment's approach C is countersigned below on the assumption of it, and says so.

A held crowd follows a beacon that moves [Proposal R13, new in r7]. The contract adopts a moving beacon on the Tram's rail, and "hold on arrival; never leave a held beacon for a farther one" does not say what holding means when the thing held to is travelling. Proposed: holding attaches to the beacon, not to the ground, so a held crowd walks after a beacon that moves and keeps holding at it. Otherwise a lit tram would shed its crowd the instant it started, which is the opposite of what EQP102's first approach — and the district that introduced the moving beacon — are built on.

Merged crowds, added in r3 at PMR004's request. Pursuit range is a property of each body, not of a crowd, so a branch holding two or three merged surges presents a contact edge proportional to the length of the occupied branch, not to the size of one surge — and a player crossing it is inside some body's pursuit range for the whole crossing rather than for a moment. Two consequences: R2's ring is what makes that legible before the player commits, and GDD §15's tuning bound should be checked against the third dispatch's merged crowd, not the first surge, because that is where the crossing is actually unsurvivable. Neither is verified; both are checkable on paper once press has a bound.

3. Crowd pressure: press gets a bound and a drawing [Proposal R2]

Adopted: a gathered crowd's edge creeps outward along its branch during the quiet interval; each surge is larger, so the held crowd's radius grows; "quiet" means a moving line you can read. Adopted tuning bound (GDD §15): on a first expedition with the panel untouched, the causeway crowd's press must not reach the causeway head before the evacuation window closes.

The problem. Press is the only thing that makes a held crowd more dangerous over time, and it has no stated rate and no stated stop. As written it is an unbounded timer, which means (a) the §15 tuning bound is a bound on an unbounded process, which is fragile, and (b) I cannot compose an encounter around a held crowd without knowing whether its edge reaches the doorway in twenty seconds or two minutes. GDD §15 already lists a cheaper alternative — "press as a fixed radius that grows on a timer rather than simulated crowding" — and REV104 has pre-registered a demand for exactly this kind of computable claim.

[Proposal R2]. Press is a crowd relaxing to a radius set by its size, not creeping indefinitely.

• A crowd holding at a beacon occupies a radius proportional to the number of bodies in it. When bodies join — a new surge arrives, or a neighbouring crowd merges — the radius grows to the new size over a few seconds. When bodies leave or die, it shrinks.
• Press is that relaxation seen from outside: after the second dispatch the crowd's edge moves outward because the crowd is bigger, not because time passed.
• A crowd with no beacon (line 4) has no centre, so it spreads along its branch without that bound. That case keeps the current behaviour, and it is the dangerous one — which is the point of §5 below.

Visible information. The forecast overlay already draws each entered crowd's current target (GDD §3). R2 adds one ring per crowd on the same overlay: the crowd's current extent, drawn on its branch. Engine-drawn, no asset. A player reading the overlay can see how much of the causeway is full before deciding whether to cross it.

Player action and cost. Killing into a held crowd now visibly shrinks its ring, which gives the cutter and the arc a purpose against a crowd nobody intends to clear: buying ten metres of road home. The cost is standing inside pursuit range to do it, and the charge spent if the arc does it.

Why this beats the alternative. The alternative is to leave press as simulated crowding and tune it. That is the one behaviour GDD §15 itself names as "the one real engineering risk", and it produces a number nobody can predict from the page. R2 makes the danger a function of a quantity the player can already see — how many came in — and it makes the §15 tuning bound checkable on paper: the first surge's radius must be smaller than the causeway. It is strictly less simulation, and it is the version I would ship first. This is a proposal to the Lead, who owns §4; I have not edited that page.

4. Charge and the once-per-organism rule [Proposals R4, R5]

Adopted: a killed organism drops a small mote of 1 unit; a mass stunned in water sheds one large mote of 5 units per 10 stunned; each organism sheds once per expedition and is visibly dimmer afterward; motes fade in seconds; walking over one banks it; a full battery takes nothing; a battery the arc is drawing takes nothing.

The once-per-organism rule is the best economic rule in the document and I am not touching it. It is what stops the harvest loop becoming a farm, and it is why the arc is a spend rather than an engine. Two things around it need work.

R4 — glow carries one bit, and roles ride on silhouette

The collision. "Visibly dimmer after shedding" puts spent-ness on brightness. The roster wants to put two more things there: a body that never glows because it does not conduct (Crust), and a body that glows brightly and unstably because it is about to discharge (Spark). That is four values on one channel, read at a glance, in a crowd the source sizes at 40 to 140. I do not believe that is readable, and I cannot test it.

[Proposal R4]. Collapse the cues where the consequences collapse.

• Glow carries remaining yield, and only that: has charge to shed, or does not. An unshed body glows; a shed body does not. A Crust glows until it is killed, like anything else — it is non-conducting, not spent.
• Role identity rides on silhouette and attitude, never on colour or brightness. The Anchor is half-submerged and still; the Grounder is on a cable, off the ground, and its position alone identifies it; the Carrier leans toward what it wants, and the forecast draws its target; the Crust is the base outline broken by encrustation; the Spark's outline flickers, which is motion, not brightness.
• Conductivity is a second, independent channel: encrustation. r1 of this page put both facts on brightness and then argued the resulting confusion was harmless; PMR004 is right that it is not — a Crust resists a stun that a spent Walker still receives, so a player who confuses them commits to a collection window that four bodies are walking through. Two facts, two channels: glow says what is left to take, encrustation says whether current reaches it. They are independent because the answers are independent.

Cost: six readable silhouettes, which is the real asset consequence of this submission (§9). Competing choice: none for the player; this is a presentation rule that makes the others fair.

R5 — a Carrier takes an uncollected mote

[Proposal R5]. A Carrier that reaches a mote on the ground consumes it and carries it; the mote is gone from the floor. Killing that Carrier drops it again as a mote. Nothing else can take a mote but the surveyor and a Carrier.

Today a dropped mote is bounded only by a fade timer, so "collect during a stun" is a stopwatch problem. R5 makes it a contest: the 20 units lying in the channel after a stun are being walked toward by the Carriers that came in with the surge. Player action: which motes, in what order, and whether to spend the burst now — which clears the collectors but empties the battery the motes were going to fill — or collect first and take contact damage. Visible information: Carriers lean toward the motes and the forecast draws their target. Consequence: charge you did not take is charge that walked away rather than charge that faded, and it is recoverable by killing the thief. Cost: you fight for a harvest you already paid a battery for.

This is also the honest answer to a fair criticism of the coil loop: at present the only things between the player and 20 units are a fade timer and the size of their battery. R5 puts an opponent there.

5. Beacon routing: what a role may and may not do to a light [Proposal R9]

Adopted and load-bearing, restated here as a limit on my own roster: the panel alone decides the intake. Current the player injects travels power, never signal; it can light a beacon steady and open a hatch, but it can never choose where the next surge enters (GDD §4). World r20 states the same as a contract on Campaign and Combat: a district may not change what power, signal or a beacon means, or make an ordinary lamp draw the Flood.

[Proposal R9], and the argument that it is inside that contract — necessary, and not sufficient. PMR005 is right that clearing the contract's prohibition list does not settle which connected consumers lose power, what alternate supply exists, or how recovery works; those are C4 in §0.4 and belong in the shared topology contract with Lead and World. What follows is the containment argument only. One role, the Grounder, can darken a beacon: it climbs a live cable run and grounds it, and the beacon on that run flickers and then goes dark until the Grounder is killed or the branch is switched.

• It does not change what a beacon means. A grounded beacon is an unpowered beacon, and an unpowered beacon draws nothing — the existing rule, applied. Dark still means dark.
• It cannot cause a dispatch, touch the signal cable, or select an intake. It removes a light; it never makes one. The panel's monopoly is untouched.
• It is announced before it acts: an irregular flicker that is visibly not the dispatch pulse — the dispatch pulse is a travelling flicker along the thin cable, panel to beacon, which GDD §4 makes the observable the whole rule's fairness rests on — plus a spark at the body's position on the thick cable, plus the forecast marking that branch failing.

Why the roster needs it. Without it, a routing decision is permanent for as long as the player leaves the panel alone, so "park the crowd and go elsewhere" is a solved problem after the first expedition. R9 puts a clock on a plan without putting a clock on the player: the plan decays because something is eating it, visibly, in a place you can go to. That is a better pressure source than a shorter timer, because it has a location, a cause and four different counterplays — kill it, re-switch, plant your own light, or take the long way — each costing something different.

6. The shared grammar — the interface with Equipment

This section exists so that EQP101 and CMB101 cannot invent two systems. I propose it as a contract and ask Equipment to accept, amend or reject it in writing. Where it constrains Equipment it constrains this page equally.

Targeting [R6]

R6 answers the question the Cutter page asks — "what must remain automatic, and what meaningful action does the player control?" If the weapon fires by itself and cannot be stopped, then what enters its reach is a decision the player is making continuously, and the target line is what turns that from an accident into a choice. It costs one engine-drawn line, and it is what makes the Spark fair: walking within reach of a Spark standing in the water you are standing in is the mistake, and the line shows you making it half a second before it lands.

A hold-fire button would defuse this entirely and I recommend against it. The contrary view is that taking control away feels bad. The answer is that the genre's premise is an automatic weapon; this design should charge for it rather than apologise for it.

Any Equipment weapon must state its targeting rule against this list: what it selects (nearest, densest, ahead of you), whether it can be withheld, and what its line looks like. A weapon that picks targets differently is a genuinely different weapon, which is what the Cutter page asks for. A weapon that does more damage is not.

Damage [R8]

[Proposal R8]: there are no damage numbers in this game.

• An enemy has a hit count — the number of cutter hits that kill it — not a health pool. The basic organism is three hits at roughly two hits a second (adopted, GDD §3).
• There are no resistances, no multipliers, no elements, no armour values. A body either conducts or does not; a body either can be reached or cannot.
• The player has five hits of health. Contact costs one, with a moment of immunity after. Unprotected energised water costs one per second. (Adopted, GDD §3.)

Why. With no XP and no weapon-damage ladder, damage numbers would be a system with one number in it and nowhere to go — and the first thing anyone would do with it is build the stat progression the Expansion Brief warns against ("a catalogue of damage bonuses alone does not solve this"). Agreeing this early is cheap; unwinding it after Equipment has thirty items priced in percentages is not.

Corrected in r3, after PMR004: this is a recommendation with a cost, not a ban, and the owner's brief forbids none of it. Different hit counts are inside R8 and always were — Campaign's Heavy takes ten, and its economic job is a good one (Enemies §0.2). What R8 argues against is a published ladder of numbers for weapons to be priced against, because that is the thing that grows on its own.

The cost of R8 is real and I will state it: it removes the easiest lever for making a late district harder, and puts the whole difficulty burden on composition, topology and the clock. §7 and the campaign arc in Enemies are my argument that this is enough. If PM or Equipment think it is not, this is the decision to have now rather than later.

Status [R7]

[Proposal R7]: three statuses for this draft, extended only by a revision of this page that states the reason.

Reworded in r3. PMR004 is right that nothing in the owner's brief bans further statuses, and that a compact grammar is a scoped proposal rather than proof the alternatives are bad. The claim is a cost, not a prohibition: each status added is another thing a player must read off a crowd of a hundred and forty, and this design spends its readability budget on where bodies are. If a fourth buys more than it costs, it should be added — through this page, so there is one list.

Chain geometry, supplied at Equipment's request (proposals). A chained hit reaches bodies within a short hop reach, proposed at three metres of the body that was hit, needing an unobstructed line between them; at most three additional bodies per direct hit, each once per hit, and a chained hit never retriggers the arc. Ties — more eligible bodies than hops — resolve nearest first, then to the body nearest the surveyor, so the chain is predictable from where you stand, which is the whole reason for keeping it bounded. Crust eligibility: a Crust is never a chain target and never a hop. It does not conduct, so the chain neither enters nor passes through it; a chain whose only continuation was through a Crust ends there. It does not absorb or reflect, and it is not a shield for the bodies behind it — the chain simply has to reach them another way, which is a positioning problem for the player.

There is no "armoured" class, and the word should be retired. Echo Relay asks Combat to confirm "ordinary/armoured classifications"; the word appears only on Equipment's pages, and under R8 there are no armour values anywhere in the design. Two real things are being reached for and they behave differently: the Crust, which does not conduct, so current never reaches it and it is neither target nor hop; and the Heavy, which conducts and is a perfectly ordinary chain target that simply takes ten hits, so one chained hit does not kill it. Written here in r14 because §12, where it first appeared at r13, has since been revised by the Visionary — correctly, and crediting it. The general statement belongs beside the rest of the chain geometry rather than inside a district-specific exception; the r13 text is in history.

The Heavy's yield exception, defined here because Equipment asked for one Combat definition. Generic rows in the effect contract describe ordinary bodies. A Heavy counts as ten bodies for a stun's yield and drops one five-unit large mote when killed rather than a one-unit kill mote. It sheds once, like everything else, and by either route: a Heavy stunned in water has shed and its later kill yields nothing, and a Heavy killed by the cutter has shed and a later stun yields nothing from it. That is the whole exception; nothing else about it differs from a Walker except its provisional ten-hit durability, which stays an explicit number (PMR005).

No slows, burns, poisons or fears in this draft. One ruling made here rather than routed: the prod's push is movement, not a status — it has no duration and nothing reads it off a body — so the prod does not make a fourth.

Charge — the interface, restated as a copy

Canonical on the Design Document; copied here so an Equipment reader has the contract in one place. Kill mote 1 unit; a stun sheds one 5-unit large mote per 10 stunned; each organism sheds once, then is spent; motes fade in seconds; a bank holds 10; a coil discharge spends a whole bank; an arc burst empties one bank in about 10 s; a bank being drawn accepts nothing; a collected mote goes to the first eligible bank in the priority the player set at a shelter (D023, adopting E3; this replaces the most-room rule r1–r6 of this page copied, and it is Inventory and Charge's to own). Plus [R5]: a Carrier may take an uncollected mote.

R3 — closed by the Lead's ruling. r1 of this page asked how long a coil discharge keeps the water energised, because the source stated the stun and not the water, so the uninsulated cost of one's own discharge could not be computed. r1 proposed one second, r2 one to two, r3 adopted Equipment's eight. The contract r3 rules two seconds of powered water and an eight-second stun, and this page and Equipment both countersign it — Equipment withdrawing its eight, this page withdrawing the eight it had adopted from Equipment. An uninsulated discharge therefore costs two health of five, and the collection window inside the stun is not gated on insulation. The row is in the effect contract at §0.2; and the branch is settled too: fixed panel power does hold that same water hot while it is lit (C2, ruled against this page, above).

Conduction and the arc — R10, withdrawn

r1 proposed that an arc chain travel through energised water. It is withdrawn; the adopted rule is Arc A1's bounded body-to-body chain, in the effect contract at §0.2. The proposal, and why it was attractive and wrong, are kept in the revision log at §12 so it can be argued again if anyone wants to.

7. Composition — the interface with Campaign

Campaign owns placement (Expansion Prompts). This page owns what a legal composition is. Proposed budget, per surge:

Difficulty is composition and topology, never numbers. A late district is harder because the water connects differently, because two roles are live at once, and because the routes home are longer — not because the same bodies take four hits. That is R8 stated as a rule Campaign can build against.

The budget against REF107 — answering X32 where the cap is actually set

EXT102's X32 now names two pages that cite REF107 zero times, and this is one of them. It is right to name it: Enemies §0.7 discusses the roster size, but the rule that caps within-district depth is the budget above, and it lives here. X32's challenge is within-district choice depth, not roster size, and this is the paragraph that limits it.

What the budget does to depth, stated as arithmetic rather than as a defence. REF107 records comparable survivorlikes carrying few environments and very large per-run variety — Halls of Torment, 6 stages against 70+ monsters; Vampire Survivors, 5 base stages — and observes that the large numbers are the things that differ within an environment. This budget caps a district at three distinct non-Walker roles and a surge at two. So the encounter variety a district can legally present is bounded by which two of its three roles are live, against its own topology and panel states — a small number of role combinations multiplied by a larger number of situations, not 70 of anything.

I am not conceding a thing the finding did not ask for, and I am not withdrawing the cap. The critic has twice noted pages answering X32 by giving away something it never challenged; the cap is a considered constraint and it stays.

The reason is legibility — and PMR008 is right that r8 stated it as a prediction, which it is not entitled to be. The honest form: my judgement is that a crowd of forty to a hundred and forty containing four kinds of body, each answering a different precedence line, would be hard to read at a glance, and that this design's central claim — that consequence can be computed before it lands — is the thing most at risk if it is. That is a design judgement about a screen nobody has seen. It is not a measurement, no page can make it one, and the number two is a first guess and not a threshold anyone has tested. Whether four kinds are legible, or six, or three, is exactly the sort of question an authorized play phase answers and prose cannot. If it turns out legible, raising the cap is a one-line change and I would take it.

What I will concede is that this is a bet with a stated cost, and where the honest lever is. Depth here has to come from situation — water topology, panel state, which routes are lit, what the player is carrying — rather than from bodies on screen, and nothing in the design phase can show that situational variety reads as richly as seventy monsters do. If the owner wants more, the lever is raising the per-district cap or placing the same eight roles against more topologies rather than adding a ninth role.

But "a combination costs nothing" was wrong, and PMR008 is right to strike it. A combination costs no new asset, which is the only sense in which it is free. It still costs a district author placing it, a shared-rule diff written for it, a written expedition checked against it, and a reviewer's per-beat pass — which on this wiki is the scarcest resource of the lot and the one every review round has actually been spent on. The accurate comparison is: a new role costs authoring plus a silhouette from a budget whose first item is still an unapproved request (Campaign §7); a new combination costs authoring alone. That is a real difference and a smaller one than r8 claimed.

REF107 proposes no count and neither does this.

A fight that yields only charge must have somewhere to spend it

Proposed as a composition rule, from X37 and PMR009 §2. The critic found a written plan whose fight "returns with charge that is discarded at extraction" against an objective the player could have walked to. PM accepted it narrowly, as a gap in a written value comparison rather than proof about a built game. The general form belongs here, because it is a rule about encounters and not about one district:

Charge never leaves the district. So an encounter whose only product is charge is worth its cost only if that expedition contains something the charge pays for — a hatch, a gate, a stun the route needs, a burst the doorway needs. A fight that yields fuel with nothing to burn it on has bought nothing, however well it is written.

What this asks of a district page, in one line: when an encounter is placed for its yield rather than for blocking a route, name what the yield buys in that same expedition, and say plainly that the fight is optional. When it is placed to block, a free bypass elsewhere is fine and usually good — the point is not to remove bypasses, it is to stop counting a fight as content when it buys nothing.

Why this belongs to combat rather than to the districts. Every one of this page's harvest interactions — the coil's mass shed, the Heavy's five-unit mote, the Carrier's stolen motes — produces the same discardable resource, so the same question applies to all of them, and it is the question the Game vision is asking when it says harvested charge should "create room to act". It is a check on the writing, not a new mechanic, and it adds nothing to the rules a player has to learn.

A district's shared-rule diff — the check REV103 r5 pre-registered for REV104 — should, for combat, name: which roles are present, which precedence line each exercises, what announces it, and what the player can do about it with the equipment available at that point in the campaign. A district that cannot fill that in has decoration, not encounters.

8. What the player actively decides while the attack is automatic

The Cutter page asks this directly, and it is the fair question to ask a game whose weapon fires itself. Under the rules above, the player is deciding five things continuously, none of them by pressing an attack button:

1. What is inside the cutter's reach — because it cannot be withheld (R6). Near a Spark in water this is the whole game; near a Crust it decides whether an arc chain has anywhere to go.
2. Which precedence line they are standing in. Inside pursuit range you are a target; outside it you are scenery and the crowd is doing something predictable. Moving in and out of that line is the moment-to-moment skill, and R2's ring is how you read it.
3. When to spend. A coil discharge needs a full battery, a crowd in connected water, and a crowd that has not shed. An arc burst needs density and a battery you are willing to empty. Both are irreversible for the expedition.
4. Where the water is, relative to both of you. Water is the weapon and the hazard on one clause. Standing in it makes the coil possible and the Spark lethal.
5. What to walk away from. Every second fighting is a second not spent on the district, and the district is the reward (GDD §6). Nothing here rewards clearing a crowd for its own sake: no XP, and spent bodies are worth nothing.

And one thing decided before the expedition: what the rack can answer. Insulation answers water. A second battery answers a big harvest. The beacon answers Carriers and Grounders both, for two slots and a battery. A rack cannot answer everything, which is the point.

9. Presentation and asset consequences — stated as a cost, not a request

No asset has been located, inspected, downloaded, purchased or created by this role, and none is authorized. GDD §14 already records the enemy sprite as the project's first likely approval request, and the Asset Licences reference records that "free" is not one thing and that tinting or resizing is an adaptation some licences forbid.

What this submission adds to that cost, honestly: the roster's readability depends on six distinguishable silhouettes, five more than the current one-enemy proposal. That is the single largest consequence of this page and PM should weigh it as such. Three mitigations, in the order I would apply them:

1. Three of the six are readable from position and attitude rather than from a new creature. A Grounder is on a cable, off the ground. An Anchor is half-submerged and not moving. A Carrier leans toward its target, which the forecast overlay already draws. If the base body can be posed, these three cost poses, not creatures.
2. Two of the remaining three are silhouette edits of the base body: the Crust is the base outline broken by encrustation; the Spark is the base outline with a flicker. Whether that is legal depends on the individual asset's licence, not its host's general statement. I claim no availability and have inspected nothing.
3. If the owner takes the humanoid option in GDD §14 decision 2 — the Flood as a humanoid mass, which is among the owner's own examples of the Flood — the roster does not change: what conducts is bodies standing in water either way, and posture-based roles work at least as well on a humanoid. This roster is neutral on decision 2 and does not push it either way.

New non-creature cues this page needs, all engine-drawn, no asset: the cutter's target line (R6); the press ring per crowd on the forecast overlay (R2); the Grounder's irregular beacon flicker, which must be distinguishable from the dispatch pulse's travelling flicker, plus a spark at the cable (R9); and a flat, mote-less feedback on a discharge into an already-spent crowd, so a wasted battery reads as wasted.

The gap I cannot close. The Spark's fairness leans on a rising audible whine as it enters cutter reach. GDD §14 says sound is not designed. Without audio the visual fallback is the target line plus the flicker, which I believe is enough but cannot claim is enough. Recorded as a gap, not as a request.

10. Open questions

Consolidated into §0.4 in r4, so there is one list rather than two that drift apart. Every question there is stated with the position this page proceeds on, as Wiki Conventions §12a asks.

11. Where the rest of this submission is

• Enemies — the six roles, why each exists, the two mixed encounters checked against the rules above, the elites-and-bosses assessment, and the early/middle/late campaign arc.
• The six role pages: Walker · Anchor · Carrier · Crust · Spark · Grounder

12. The Choir Cascade — finite environmental exception

2026-09-07 · Game Visionary disposition after PMR011. Combat r13 / 600893 correctly showed that ordinary Arc damage would make the flagship build weaker than its parts and correctly separated Crust from Heavy. Its four-origin/sixteen-body answer is superseded because it explicitly did not deliver the owner's screen-wipe target. The full prior reasoning remains in history.

Current rule

• A tuned Choir network is a finite, visibly outlined wet footprint with seven connected bells and a master bell. It is an environmental discovery, not an Echo Relay item.
• The player brings Coil, Arc and two separately linked full Batteries. Firing both through the master bell empties both banks and trips the network dark for the rest of the expedition.
• Coil first stuns conducting bodies in the footprint. The network then gives each stunned ordinary body in that footprint one Choir strike; that body dies. A body is claimed once. Because the whole marked footprint fires together, no origin order or four-/seven-origin subtotal is needed.
• This lethal result is Choir-specific. Away from a tuned network, Coil and Arc retain §§0.2/6: Arc chains body to body to at most three additional targets per direct hit, and a chained hit contributes ordinary Cutter damage even when the target is stunned.
• Crust is untouched because it does not conduct. A Heavy conducts and takes one strike, not the ten hits needed to kill it. Dry bodies, bodies outside the lit footprint and a later wave survive.

Why this exception earns its place

Without the network, the same five-cell rack can stun the basin and use standard Arc while the player fights, but it cannot erase the room. With the network, prior observation, a cargo-heavy loadout, two full banks, crowd positioning and a one-use authored place convert familiar tools into the promised ridiculous build. The visible basin is both preview and numerical bound; clearing every eligible body inside it is bounded propagation, not a global rule or an unbounded chain. Whether the crowd size, readability, timing, performance and aftermath feel spectacular remains untested.

13. The vision's combat example, checked against the rules

PMU008 asks domain designers to answer for their systems on the Visionary's actual changes. The Game vision's "Why combat belongs" paragraph is a combat example, so this role checks it. Its design intent is right and this page endorses it without qualification: "Combat earns its place when it buys an intervention rather than replacing one… Fighting with no useful place to spend its local charge does not earn a required encounter." That is §7's composition rule in one sentence, arrived at independently, and it is a better sentence than mine.

The scene it summarises is correct. Campaign r9 §3.1's Lock Flight beats at 3:30 and 5:00 check against the shared rules: ten walkers taken at the crowd's edge inside pursuit range, ten motes walked over as they fall, two hits taken, one bank filled; then a discharge at the head basin's edge that opens the lock-house door for its two seconds. Nothing below is a problem with the design — all three are things the summary drops that a reader will need.

Why the third one matters more than the others, and why I would not let it stand. As written, the owner reads that a coil discharge powers what you aim it at and nothing else. The truth is the opposite and it is the better story: a discharge lights everything on the wet circuit, and the surveyor does it anyway because they know the crowd is held at a nearer light and cannot answer this one. That is the vision's own signature moment — "I understand why this place is dangerous. I can change it." — happening in one sentence. The correct rule is more impressive than the incorrect summary of it, and it costs the paragraph four words.

Ownership. The Brief is the Visionary's page and this role has not edited it; these are three offered clauses, all lifted from Campaign r9's own beats rather than invented, and the Visionary and PM own whether to take them. No rule on this page changes, and nothing here asks for a new encounter, role or cost.

14. EQP102 countersigned — the three approaches against targeting, press and pursuit

CMB102 asks this role to countersign or revise EQP102's three approaches to one Tram objective. Countersigned, with two rules this page had to supply and one interaction Equipment did not claim. This is a check of the combat grammar only: placement is Campaign's, the items are Equipment's, and nothing here is a judgement of whether the three are equally attractive, which no page can establish.

One interaction Equipment did not claim, and could have. A planted beacon carries a bank, and a bank with units in it is loose charge — so under R5 and the Carrier's role target, a planted lure draws Carriers as well as light-followers, by a different clause. They walk to it because it is charge, not because it is lit, and they keep walking to it after the light dies for as long as the bank has units. In a district that fields Carriers, C's lure is therefore doing two jobs and outlasts its own light for one of them. That is a genuine argument in C's favour that its author did not make.

What I am not countersigning, because it is not mine: the one-unit-per-ten-seconds lure burn, the acquisition chain, the cell counts, and whether the three approaches are comparably attractive. Equipment and Campaign own those, and the last is unverifiable in design.

15. Revision log — what changed, and what was withdrawn

Kept short and at the foot, so §0 is the current contract and this is the history. Full text of every revision is in page history.

The withdrawn R10, retained because it may be argued again. The proposal was that while a body of connected water is energised, an arc chain propagates through that water as well as body to body, reaching every conducting body the water connects. It was attractive because it made the late payoff cost five separate preparations — the arc, a full bank, Sparks deliberately left alive, a crowd routed into connected water, and insulation or dry footing. It was also a screen-clear with a story attached, which is what PM saw and I did not. If it returns it needs its own argued change; it is not a property of electricity that clause 1 implies.