# Ascension: Before the Ash — game design draft

Version 0.3 • September 23, 2026 • Working title and original additions subject to canon review.

## The game we are making

You play Pyrrhus Menethil before he becomes the figure remembered by the Dawnbreaker chronicle. Exiled with his mother as the king's firstborn illegitimate son, he has learned to protect people without expecting authority to protect him. Civil war gives him the means to change that—and opportunities to become the kind of ruler who once discarded him.

The game combines intimate, height-based tactical battles with a compact provincial campaign. Pyrrhus physically travels with his company. Away from him, governors, commanders, warlords, and his brothers continue making decisions. A captured bridge changes a supply route. A conscripted harvest changes winter. A beloved companion left governing a province cannot also fight beside him tomorrow.

The main emotional question is: **What do you owe the people you cannot personally save?**

The references supply different ingredients: FFT for readable tactical spaces, individual turns, class combinations, and political intimacy; Gemfire for provinces, commanders, resources, and conflicts that continue beyond the protagonist; Ragnarok Online for expressive sprite silhouettes and appealing costume identities; anime portraiture for faces and dramatic scene composition. These are design influences, not a requirement to reproduce their exact systems, characters, or interface.

### Five pillars

1. **Every victory has a home address.** Tactical outcomes change a named settlement, resource, route, or relationship.
2. **Build people, then trust them.** Developing a companion improves both their tactical role and their usefulness away from the party.
3. **Power is an obligation.** A new province adds duties and political promises alongside income.
4. **Pressure is readable.** Warnings, routes, travel time, and likely losses are visible before committing.
5. **Warmth makes loss matter.** Meals, jokes, festivals, recovery, and affection receive space between crises.

## The player's rhythm

An ordinary session: read a dispatch; discuss the next move in camp; adjust jobs; leave orders with a steward; choose a route; enter one substantial tactical battle; see its local aftermath; share a short character scene. Aim for 60–70% combat and preparation, 15–20% travel/governance, and the remainder narrative. These are pacing targets to test, not rigid quotas.

At each layer the player should be answering one question:

| Layer | Main decision | What it changes elsewhere |
|---|---|---|
| Company | Who travels with Pyrrhus? | Available commanders and governors |
| Battle | What must survive, and what can be sacrificed? | Infrastructure, prisoners, recruits, public trust |
| Travel | Which crisis gets the hero's attention? | Arrival dates, provisions, intelligence |
| Province | Which promise gets funded this week? | Recruitment, unrest, supply, alliances |
| Story | What kind of authority will Pyrrhus accept? | Companions, political settlement, ending |

## Cinematic combat and the first playable

**Creator-confirmed direction — September 23, 2026.** Ascension uses compact sprites on the tactical map and larger, expressive side-view sprites for automatic fight scenes. Street Fighter and Guilty Gear are references for pose clarity, anticipation, impact and dramatic framing. The characters are original Ascension designs; these references do not imply borrowed assets or fighting-game controls.

[Play The Old Road prototype](prototype/). One generic company swordsman faces one crimson sentry on a small isometric field. This is the first working foundation for the combat loop, not a complete campaign slice.

### The loop

1. Read the battlefield, terrain, remaining HP and movement range.
2. Move the selected unit, choose a target and review the attack forecast.
3. Confirm the action on the tactical map.
4. Watch an automatic side-view exchange: ready pose, approach, sword swing, hit reaction, damage number and HP loss.
5. Return to the same tactical positions with the result applied exactly once. The enemy takes its turn; repeat until the encounter ends.

**There is no player interaction during the fight scene.** No QTEs, timed strikes, parry inputs, combo execution or mid-scene defensive choices. The characters visibly attack and react to one another. Player agency belongs to tactical decisions before an action is committed. Any future skip or fast-forward control changes presentation duration only, never the resolved outcome.

### Two coordinated sprite scales

Map units have compact, approximately three-head proportions. Their idle is the **same directional walking animation played in place**, at a calmer cadence; movement advances that cycle along the route. Do not substitute a separate breathing idle, squash/stretch or a bouncing still image. Keep the foot anchor and costume identity stable across facings.

Cinematic fighters have larger, more articulated silhouettes, readable weapons, and distinct ready, attack and hit-reaction clips. Preserve the same hair, armor, palette, identity and equipment as the map unit. Stage both fighters together so the strike connects with a visible reaction. Camera movement, hit effects and damage numbers support the animation; they cannot replace it.

AutoSprite supplies the character map and cinematic sheets for this prototype. Terrain and interface are authored in code. The map exports contain sixteen walking frames per direction at 192 × 192 source resolution. Cinematic source cells are 384 × 384; ready clips have 24 frames, replacement attack/hit clips have 32 source frames, with unsuitable motion trimmed in playback. These are generated pilot assets, not hand-cleaned native 64-pixel production masters.

### Implemented rules for The Old Road

| Rule | Current prototype |
|---|---|
| Battlefield | 8 × 7 cells; grass and road passable; water, trees and rocks blocked |
| Player | One swordsman: 32 HP, 10 attack, 3 defense, 4 movement |
| Enemy | One sentry: 24 HP, 9 attack, 2 defense, 3 movement |
| Turn | Player turn followed by automatic enemy turn; one move, then attack or wait |
| Sword reach | One orthogonally adjacent cell |
| Damage | Attack minus defense, minimum one, clamped to remaining HP |
| Accuracy | 100%; hero deals 8, enemy deals 6 before lethal clamping |
| Reactions | Defender visually recoils; no automatic counterattack inside the same scene |
| Outcome | Win when the sentry reaches zero HP; lose when the hero reaches zero HP; restart resets the encounter |
| Replay | Replays the last recorded exchange without changing HP, position or turn |

Attack ends the activation. The enemy independently moves and attacks on its turn. The broader Readiness timeline, action-before-move, rear bonuses, guard, skills, height, rescue windows and campaign consequences remain later systems; none are silently simulated in this small test. Zero HP ends this test rather than invoking the proposed campaign wound rules. A crimson generic opponent is a test faction marker, not a revision to Whitebridge's story allegiances.

### Foundation and acceptance

Encounter data, pure battle rules, the turn controller, sprite metadata and rendering are separate modules. The forecast and resolution share the same calculation. A resolved combat event carries attacker, defender, damage and before/after HP to the presentation layer; animation never rolls or reapplies damage.

The first implementation is dependency-free browser JavaScript and Canvas, playable directly from this bible. Godot remains the proposed production engine for a larger terrain and campaign build; porting the rule contracts and content requires a deliberate integration milestone. This browser prototype does not claim a Godot project already exists.

The test is complete when a player can move, watch both sides attack, observe persistent damage, win or lose, restart and replay safely. Production readiness additionally requires tighter weapon continuity, stable anatomy and foot registration, approved map-to-fighter identity, authored defeat/cast/skill clips, sound design, and performance testing at full encounter scale. Equipment is baked into these pilot sheets; modular weapon swapping remains future work.

The earlier low-poly/AI 3D close-up studies and defensive-response mockups are **superseded** by this decision. They remain archived visual research. They do not define the current gameplay contract.

## Tactical combat

**Scope:** The following describes the intended full game. See [the first playable](gameplay-cinematic-combat-and-the-first-playable.html) for its deliberately smaller implemented rule set.

### Battle structure

Deploy six controllable units initially, rising to eight for major battles; maintain a roster of roughly 16–24. Named allies and generic recruits follow the same core rules. Story guests occupy explicitly shown guest slots. Ordinary enemies number 6–12, with reinforcements used to create pressure instead of enormous starting crowds.

Maps begin around 12 × 12 logical cells; later maps may reach 20 × 20. A cell can contain multiple distinct walkable surfaces, such as a bridge over a road. Height, stairs, cover, facing, and terrain create the puzzle. A 15–30 minute ordinary battle is the initial target. Larger climaxes must justify their duration through changing objectives.

Use an individual readiness timeline. A unit activates at 100 readiness, may move once and take one action in either order, then chooses facing. Unused movement is lost; partial move–act–move is an explicit advanced skill rather than a universal rule. Actions incur recovery, so light actions cycle faster than heavy ones. Charged abilities put their resolution on the timeline and preview whether targets can move first. A bounded speed range prevents permanent turn monopolies.

Call this meter **Readiness** in this game: the later tabletop site uses AP in a way that should not be silently copied into a different action economy.

### Terrain and positioning

- Four logical facing directions; rear attacks gain a clearly previewed bonus. Later tactical prototype proposal: +20% HP damage from directly behind, no diagonal rear bonus, borrowing an existing setting rule with explicit adaptation.
- Height improves sightlines and some ranged attacks, but every ability specifies its climb and vertical limits. A roof is valuable because of approach geometry, not a universal damage multiplier.
- Low walls obstruct movement and reduce some ranged accuracy. Full obstacles block line of sight. Arrows and direct spells can use different trajectory tests.
- Water, mud, stairs, doors, ladders, and broken bridges have distinct movement costs or links. Fire spreads only over authored combustible regions with a visible next-step preview.
- Units can protect an exit, intercept a carrier, shove an enemy off a safe ledge, or occupy a signal point. Fatal falls are rare, marked, and accounted for in the attack forecast.
- Clear allied passage rules: move through allies, never end inside them. Enemy zones of control stop ordinary movement; mobility jobs have explicit exceptions.

### Forecast and randomness

Before confirmation show hit probability, HP damage, guard damage, status chance, collateral effects, retaliation, and affected units. Prototype HP damage is a simple attack-minus-defense calculation, minimum one on a successful damaging hit. Defense is separated into physical and magical resistance. Armor also has a finite **Guard** meter: Break skills damage Guard, and Guard at zero exposes the target for a short window. Avoid making armor simultaneously an endlessly regenerating health bar and a hidden multiplicative reduction.

The existing tabletop rules use an effectiveness roll after calculating damage. For this digital game, test a simpler hit model against that roll; select one before balancing content. Do not stack both sources of variance by default. Use seeded randomness, honest percentages, and bounded outcomes. The displayed forecast must use the same rule functions as resolution.

### Morale, wounds, and recovery

**Company cohesion** is the campaign measure of trust in leadership. **Unit resolve** is a local battle resource influenced by nearby allies, casualties, personal obligations, and rally abilities. **Province unrest** belongs to civilians and is separate. The player should never have to infer which of three identically named morale bars a decision changes.

Low resolve first weakens reactions and resistance to intimidation. At zero, ordinary enemies can surrender or withdraw; the player's units do not randomly discard a turn. A player can choose an emergency retreat, or spend a command resource to hold a vulnerable formation together. Visible traits can give a companion a personal crisis, but it is a narrative event with warning rather than an unexplained AI override.

At zero HP, a unit becomes downed with a three-activation rescue window shown on the timeline. Stabilization, carrying, or winning with control of the body avoids death. A recovered unit gains a wound requiring rest, treatment, or a temporary assignment away from combat. On standard difficulty, an unrecovered unit is captured or suffers a severe injury where the fiction supports it; explicit lethal conditions remain possible and are signposted. On optional Iron Chronicle difficulty, an expired rescue window can cause permanent death. Pyrrhus's death means battle failure, not an alternate campaign without the protagonist.

Other mandatory story characters receive consistent capture/injury outcomes on standard mode; optional permadeath requires fallback scenes authored before that mode ships. No Academy recall beacon technology is assumed for this historical period.

## Jobs and company building

Character level provides broad competence. Job rank unlocks abilities with earned JP. A unit equips one primary command set, one learned secondary command set, one reaction, one support trait, and one movement trait. Give the UI a complete explanation of job requirements. A character can change jobs freely in camp; field changes require returning to a safe location. Learned abilities persist.

Avoid permanent stat-growth traps: long-term character stats come mainly from character level and equipment, while the current job supplies a reversible modifier. A player experimenting with a healer should not ruin a swordsman. Ordinary equipment has strong identities and small upgrade ladders; avoid a constant flood of randomized loot.

### Proposed launch job roster: twelve shared jobs

The later Academy class tree is canon for its own period. This proposal uses historical battlefield professions; it does not assert that future institutions already exist.

| Job | Tactical identity | Sample ability | Campaign use |
|---|---|---|---|
| Vanguard | Shield positioning and safe approaches | Interpose: protect adjacent ally | Train defensive militia |
| Archer | Elevation and lane control | Pinning Volley: restrict movement | Patrol roads |
| Scout | Mobility, flank attacks, observation | Mark Route: safe traversal | Reveal raid routes |
| Apothecary | Finite supplies, stabilization | Field Dressing: stabilize and remove bleed | Reduce recovery time |
| Acolyte | Controlled elemental offense | Ember Line: narrow delayed strike | Evaluate local elemental hazards |
| Sage | Protection and sustained recovery | Hearth Ward: temporary safe area | Improve recovery camps |
| Knight | Guard breaking and holding ground | Shield Break: damage Guard | Fortification inspection |
| Lancer | Reach and positional commitment | Brace: punish approach | Train anti-charge levies |
| Engineer | Structures, bridges, siege devices | Field Ramp: temporary traversal | Repair roads and mills |
| Standard Bearer | Resolve and formation support | Rally Standard: area resolve recovery | Reduce desertion |
| Spellblade | Weapon magic and costly burst | Tempered Edge: elemental next strike | Train mixed detachments |
| Envoy | Surrender, support, misdirection | Terms of Mercy: conditional surrender | Negotiate passage and prisoners |

Pyrrhus starts with a personal **Wayfarer** command set alongside ordinary jobs. Later **Oathbearer** abilities reflect relationships and hard-earned command, not proof that his blood makes him superior. Named companions get a small personal ability package rather than twelve additional bespoke jobs.

Example builds: a Knight with Apothecary rescue tools holds a corridor; an Archer with Scout movement reaches the mill roof; a Standard Bearer with Sage wards steadies a refugee escort; a Spellblade with Vanguard protection trades burst damage for an endangered ally's survival.

Recruitment comes from villages, freed captives, surrendered enemies, and contracts. Every recruit has a name, home province, one background trait, and a short personal note. Losing their home alters dialogue and willingness to accept harsh orders. Do not generate sprawling procedural biographies that the game cannot meaningfully acknowledge.

### Equipment and progression boundaries

Five equipment slots: weapon, off-hand, armor, accessory, and supply kit. Two-handed weapons occupy the off-hand. Consumables draw from the equipped kit during combat and national supplies when replenishing at camp. Cosmetic armor variants are limited by job family; every item does not require a new body animation set. Personal keepsakes can carry narrative identity without occupying the strongest combat slot.

Earn most experience and JP from completing objectives, participating, and rescuing allies—not repeatedly striking a friendly unit. Training catches reserves up to a moving floor below the company median. Optional contracts consume campaign days and have finite rewards. This supports experimentation without an infinite grinding loop that makes war pressure meaningless.

## The country and its armies

### Scale

Use **12 provinces** in the intended full campaign, with 3 in the first playable slice. Initially Pyrrhus commands one company and has stewardship responsibilities for one refuge. Authority expands through negotiated charters, occupations, and alliances. He need not personally annex every province to finish the story.

Illustrative provinces: Cinderford (river crossing and grain), Greyfen (wetland villages), Whitebridge (truce road), Thorn March (disputed eastern frontier), Saltmere (port and customs), Frostgate (northern pass), Bellwether (orchards), Highmere (noble estates), Emberfield (smiths), Redwater (irrigation), Saint's Rest (hospitals and archives), and the Crown Demesne (royal seat). All are new working names and their geography is deliberately unbound to the canonical atlas.

### The ledger

Country resources: **coin**, **provisions**, and **available manpower**. Province values: food stock, public order, garrison, war damage, and allegiance. Food stock feeds provision shipments but remains physically located; a safe road matters. Present public order as 100 minus unrest, not a second independent simulation variable. War damage reduces output and creates reconstruction needs.

One provision unit is an abstract shipment, not one loaf. A local food stock is reported in the same units. Manpower is available recruits, not the whole civilian population. The simulation tracks displaced households separately for relevant events rather than pretending that all humans are interchangeable recruitment currency.

Proposed early-week economy for a connected province:

- Food produced = base yield × season factor × (1 − war damage / 100), rounded down.
- Subtract civilian demand and local garrison demand first. Allocate any export as an explicit order; never silently starve locals to refill the hero's packs.
- A 100-soldier garrison consumes 10 supply units each week in the first balance pass; Pyrrhus's traveling company consumes 1 unit per travel day, including animals, attendants, and transport losses. These ratios need simulation against actual force sizes.
- Public order changes through security, shortages, requisition, promises kept, and occupation. At unrest 60 show organized resistance; at 80 warn of a possible revolt with a response window. Neither is an instant game-over switch.
- Coin pays wages, repairs, purchases, and bribes. Manpower replaces losses and fills new detachments. Conscripting during harvest reduces next week's yield and increases unrest.

These values are test scaffolding. A spreadsheet or simulation run must validate them before treating them as tuned rules.

### Orders and delegation

Each province can maintain one civil project and one military stance. Available civil projects: repair, distribute relief, collect ordinary taxes, requisition, resettle, recruit, negotiate. Military stances: defend, patrol, escort, raid, muster. Regional templates let the player set a policy once and respond to exceptions. Distant orders arrive through visible courier delays, initially one day per road segment, while trusted stewards continue standing orders.

A named officer can be assigned as steward or field commander. Their command, stewardship, diplomacy, and conviction affect outcomes. Their job gives a modest contextual advantage. They are unavailable to Pyrrhus's battle party until recalled and returned by travel. Administrative skill should not force the strongest combat unit into permanent desk duty; train lieutenants and allow a modest bench of specialist noncombat officers.

Examples: Yselle improves tax predictability and fort defenses but may prioritize contracts over popular relief. Mira reduces sickness but will resist discriminatory hospital access. Seris can negotiate passage in the marches but cannot simply order independent clans around. They tell Pyrrhus what they intend to do before it becomes a surprise betrayal.

### A clock that creates choices without rushing menus

Time advances only when the player commits travel, resting, a project wait, or a mission with an explicit time cost. Reading dialogue, inspecting maps, building a character, and considering council choices are free. A tactical battle consumes its stated campaign duration once, independent of real-world playtime.

Advance the campaign one day at a time internally. Resolve dated raids on their due day; on every seventh day resolve economy, wages, project progress, and standing policies in a documented order. If a five-day trip crosses an event, pause at the last reachable decision point and explain available responses. Route preview lists every known event crossed, provisions required, and safe alternatives.

Critical incursions are normally telegraphed 3–5 days ahead. A blind border may produce shorter warning, but the interface must identify that intelligence risk. Named deadline missions state the failure outcome before departure; the player can choose a loss rather than discover a silent campaign lockout.

### Battles away from Pyrrhus

Territorial forces are detachments with soldiers, readiness, equipment quality, commander, provisions, and morale. They are distinct from the six-to-eight named units on a tactical board. A tactical squad represents the decisive local action within a larger engagement; a single sprite does not ambiguously switch between one person and an entire regiment.

Remote clashes resolve through a transparent model: relative soldiers × readiness × equipment, modified by terrain, supply, command, and stance. Preview outcome bands such as "likely hold; 10–25% losses" with intelligence confidence. Store the random seed at commitment. A log explains meaningful factors after resolution. Avoid an opaque power number that lies about captives and supply losses.

The player chooses a doctrine—hold, delay, withdraw, or costly counterattack—and can dispatch reinforcements or negotiate beforehand. Pyrrhus can personally intervene only if he can arrive. Do not let him teleport into every remote battle. A future feature may let players command a lieutenant in an authored battle, but that is outside the first release scope unless playtesting demands it.

Wars between third parties can cut trade routes, send refugees, change access, or create a coalition opportunity. Each faction gets a limited operational budget per week and a coherent agenda. It must pay supplies and obey travel rules. Early story chapters constrain the simulation to preserve a learnable situation; later chapters relax those constraints, transparently marking ceasefires and mustering periods in the fiction.

### Defeat without a campaign soft lock

Losing a province removes its income and recruitment, creates displaced households, and changes access. It does not silently delete a required story character or the only route to the next chapter. If Pyrrhus loses every holding, the company can operate from an allied refuge or mobile camp with stricter supply limits and recovery missions. A country devastated by neglect reaches a visibly worse political settlement even if the hero survives his personal campaign. Optional challenge settings may make national collapse a formal defeat; standard play favors recoverable setbacks and a difficult ending over an unwinnable save.

At every chapter transition, validate that the remaining faction relationships, travel graph, and required witnesses still permit at least one authored continuation. Escape routes, substitute witnesses, and prisoner exchanges are designed content, not emergency rewrites triggered by an AI narrator.

## The consequences must cross layers

| Tactical outcome | Immediate effect | Later consequence |
|---|---|---|
| Save Cinderford's mill | Extra objective completed | Restored supply and cheaper relief |
| Burn the bridge to stop pursuit | Fewer pursuers | Trade detour until rebuilt; refugees need boats |
| Accept a garrison's surrender | Fewer casualties | Prisoner provisioning and negotiation leverage |
| Rescue a rival's officer | Lose time securing treasure | Unlock a future ceasefire channel |
| Loot village stores | Refill company provisions | Hunger, recruitment loss, companion objection |
| Lose a granary while winning | Enemy defeated | Winter shortage remains a real problem |
| Hold the road but lose your captain | Route stays open | Weaker remote defense, recovery/capture mission |

The game should often say **"Victory, at a cost"**, but never use a hidden cutscene to erase a well-executed optional rescue.

### A worked campaign dilemma

Day 12: Pyrrhus is at Cinderford. Whitebridge is three days away and the truce begins on day 16. Scouts report a raid reaching Cinderford on day 16. The mill is broken. You have enough funds to repair it or hire reinforcements, not both.

1. Travel now, leave Yselle defending, repair the mill. You meet the brothers and restore the food route, but fight the next mission without your strategist; Cinderford risks bounded garrison losses.
2. Hire reinforcements and travel with Yselle. Defense improves and the party remains strong, but the delayed mill means a later relief obligation.
3. Stay and lead the defense. You protect the town personally; a dispatched envoy can preserve part of the truce, but Ciro believes you rejected his invitation and Xeno's evidence travels through a less secure route.

There is no free fourth option created by opening and closing menus. There can be a better prepared outcome if earlier diplomacy earned a neighboring patrol or you preserved the reserve funds. Preparation should matter.

## Travel, camp, and everyday life

Travel uses a node-and-road map with an animated company marker, not a huge open world. A route can be safe but slow, direct but exposed, or costly by boat. Scouting reveals enemy movement and an alternate approach to the next tactical map. Encounters are curated and limited; avoid repeated random battles on every familiar road.

Camp contains treatment, job training, letters, council orders, and optional conversations. Towns are small explorable hubs or illustrated location menus with a few isometric interiors. The first slice uses one camp and one town hub. As war damage rises, the same hub loses market stalls or gains refugee shelters; recovery brings them back. Reuse communicates history.

Bond scenes include all close companions, not only romances. Remember one concrete shared detail—who fixed a boot, shared a last orange, or taught a song—and return to it later. Letters from governed provinces show consequences through people instead of only meters.

## Interface and accessibility

- **Combat:** central unobstructed board; turn timeline above; selected unit below left; action and forecast at right; objective always visible. Threat overlays have patterns as well as colors. Camera cutaways remove occluding roofs.
- **World:** travel route and event calendar on the same screen. Toggle political, food, threat, and damage overlays instead of displaying all at once.
- **Council:** show current state, chosen order, cost, arrival/completion date, and predicted change together. Batch routine orders; promote exceptions.
- **Company:** portrait, sprite, job slots, equipment, and learned abilities; compare a proposed loadout without committing. Job unlocks show exact missing requirements.
- **Dialogue:** adjustable font, backlog, name labels, optional portrait animation, and explicit consequence wording where Pyrrhus could reasonably know the result. Distinguish a flirt from ordinary kindness.
- **Chronicle:** separate witnessed facts from rumors. Track promises, deadlines, casualties, and relationship memories. Avoid a moral good/evil score.

Support keyboard/mouse and controller from the slice onward. Include remapping, scalable UI, reduced flashes, reduced camera motion, instant text, animation speed controls, and save-anywhere outside an active resolution. At a 1280 × 720 reference interface, target body text of at least 18–20 screen pixels; gameplay art can have a different internal resolution. No critical information depends on hearing, color, or a timed dialogue response.

## Scope and success criteria

The full vision is roughly 25–35 hours for a first route: 24 main battles, 6–10 optional battles assembled from a limited map kit, 12 provinces, 12 shared jobs, about 8 core named companions, and three romance arcs. This is a multi-year content undertaking for a small team, not a reasonable first milestone. Route differences mostly change objectives, allies, conversations, and endings on shared maps; they should not triple the campaign.

The first playable slice is 60–90 minutes: three provinces, two battles, four shared jobs plus Pyrrhus's personal command, one remote clash, one wound, one governance cycle, the truce reunion, and an early Mira bond scene. It must prove that a combat decision changes a later campaign decision in a way players notice.

Exclude multiplayer, procedural continents, free-roaming cities, voiced full dialogue, generational dynasties, huge armies represented by hundreds of agents, and a complete Purge campaign. The full game's ending is a local political resolution, not the catastrophe or the Dawnbreaker pilgrimage.