EchoSpire: Portals — Suggestions¶

Status: Point-in-time assessment, 2026-08-28. Not canon and not a plan. Provenance: Grok 4.6. Reassessment requested via grok-eval-instructions.md. This file is the leftovers: sequencing, first-game craft, Kickstarter presentation, and things that did not belong in the concept/story/architecture/assets documents. Verification: Same pass as the four companion evaluations. Nothing here is a backlog item until it is copied into docs/official/todo-inventory.md by you. Known errors: none at publication.

Date: 2026-08-28 Companions: 2026-08-28-overall-concept.md, game-story.md, game-architecture.md, game-assets.md

Evaluations cannot decide anything. If a line below is useful, it becomes a todo-inventory row or a bible decision. If it is not copied, it is just what an outside reader thought on this date.

Settled traps this file will not reopen: WPF is the shipping client; Unity is later; SQL Server + Kusto; House names are Axiom / Syndicate / Censors; protagonist is Alden; cards.json is the card list; _archive/ is not canon.

The only sequencing that matches the concept¶

Do the next six months as if Kickstarter were a demo appointment, not a world premiere.

Slice (ship this, film this, fund this):

Two Houses, two classes. Suggestion: Valerii Anchor (the wall, readable in five seconds) and Salvari Machinist or Drifter (the grease, the hidden door). That pair sells the thesis: compliance vs the people downstairs. Syndicate self-harm and Catalyst mutate are harder to teach in a trailer.
One three-realm Quick Play that cannot fall back to Strike.
Starters that fire the class verb on turn one.
Opening as the official lie, short, playable, skippable only after the first conflicting-order choice so Skip cannot erase the trick.
Trailer: 40s propaganda (already have a shape) + 50s of this UI playing those cards. Same House names in both halves.
Display-name mapping so the demo does not say Syntacta.

Not in the slice: Unity, 5×5, hero levels 2–10, rival heroes, consumable layer, five campaign novels, a second admin, more telemetry writers, filling all 108 empty cards across every class.

The overview already says this. The failure mode is knowing it and not doing it.

First game, no writer, no artist¶

This is survivable. The pattern that kills it is volume as comfort.

You write code for a living. Generated cards, generated plates, generated lore, and a growing platform all feel like work. Kickstarter backers do not buy work. They buy a specific hour.

Practical rules:

You are the commissioning editor. story/claude.md is already the job description. Apply the same cover test to cards and UI strings. If you would not ship a PR that failed tests, do not ship a card whose text does not execute.
Hire taste, not labor, if you hire at all. A few days of a deckbuilder designer on the two-class starter packs, or a writer to pass the orientation against the voice sheets, will move the needle more than another 80 AI cards. You do not need a studio.
AI art is allowed if it is consistent and you do not apologize in the trailer. You will get comments. A clean combat GIF is the answer. A manifesto about tools is not.
Do not write novels in the bible and then skip Phase 2. The next story object is production copy for one orientation and one induction, not Session 6 cosmology.
Play the game. The harness is for after the verbs work. Sit at WPF, pick one combo, lose on purpose, notice whether you were asking the class question. If you were playing STS-with-renames, the pack is wrong.

Kickstarter presentation¶

Say Windows / WPF out loud. “Built in a custom engine, Windows first, console-quality rules, Unity presentation later if we fund it” is honest. Silence will be read as “they couldn’t do Unity.”
Do not show the 15-panel truth dump. It spoils the only trick and violates your own rules.
Do not show ART PENDING, mixed House names, or the snail boss (if that fixture still exists in any gallery path). Dev gallery is not a trailer source.
One sentence on the page: “You play the official history of a broken world. Then you find the records nobody thought to forge.”
Hide Salvari in marketing, as already ruled. The fifth House is a discovery, not a bullet.
Proof of ship: tests, a playable build, a content pipeline. Engineers on Kickstarter respect this. Everyone else needs the GIF. Lead with the GIF; put the engineering in Updates.
Goal vs 5×5. Price the campaign as two Houses and a finishable 1.0, with the rest as updates. A funded 5×5 you cannot finish is how these pages become cautionary tales.

Cold viewers already failed to read the teaser as propaganda. Assume that. The demo has to be the broadcast, then break.

Product habits that will leak¶

These are small and expensive if ignored:

Default appsettings auto-login as admin and SkipOpeningStory. Ship config is a product surface.
CORS AllowAnyOrigin() before any public API.
House vs faction in UI chrome. The bible ruled this. A “Choose your House” screen that says Faction in the header is a tell.
Quest-beat navigate with no view. Silent no-op. Either wire it or stop calling it.
Two loops. WPF Quick Play is canonical for the slice. Console/Sim prove Core. Do not implement the same shop twice.
Promote-on-boot in Development. Until a fight cannot see the bootstrap 49, you will keep evaluating the wrong game.
Do not print the graphic novel captions as a stretch reward until they match the Wilson rule and the pillars.

What I would not do¶

Relitigate Unity, Postgres, House names, Alden’s name, or “should Core own combat?”
Grow the card pool before the 108 and the starter verbs are honest.
Build an effect-module marketplace, a public community API, or multiplayer.
Commission five faction anthems before one combat loop is fun.
Write the player as a blank for twenty hours. Negative space is a novel luxury.
Balance with RandomPolicy and then trust the dashboard.
Start Campaign Act 2 outlines for Houses that are not in the slice.

If the question is “can this work?”¶

Yes, as a funded, finishable, respected indie roguelike with a real thesis — if the next visible object is a specific good hour.

No, as a 5×5 platform launch on the current teaser plus a catalog whose identity cards do not execute.

You already have the rare half: a tested engine, a client that plays, a story discipline document, and a catalog that is large. The common half is still due: a first hand that is obviously this game, an opening that is the lie, and the willingness to ship less than the engine allows.

That is the whole suggestion.