Unofficial concept
Support · route control
The WaywardenGlimmerrootWorking theme · Grove
Unofficial concept
Support · route control
Playable tile concept · original & unofficial
Glimmerroot
the Waywarden
A finished character pitch you can understand in under a minute: expressive tile art, a playable hex-board movement mechanic, and a complete downloadable handoff for Connor.
What it isA route-control support hero
How it playsPlace a shortcut both teams can contest
What shipsTile pack, rules, browser prototype, and 3D source
Concept mockup—not an official build. No Spell Valley logo, interface art, character assets, or private production files were used.
LIVEplayable hex demo
5tile export sizes
33D animation clips
PASSGLB + FBX re-import
Playable mechanic · guided prototype
Place the light. Race for the route.
Drag the suggested piece onto a glowing legal hex—or tap the piece and destination. The board accepts only the move the current rules allow.
Step1 / 3
Do this now
Drag Waylight onto the hex labeled “Drop light”
The board dims everything except the active piece, its route, and the only legal destination.
Drag or tap
ObjectiveReach the gold hex this turn
Your turn
Concept board · demo assumptions
Active abilityWaylight
Prototype cost · TBD
Drag this marker onto the glowing hex.
Legal moves onlyInvalid drops snap back and explain why.
Real hex logicDestinations come from axial-coordinate adjacency, not visual guesswork.
Every input coveredDrag, touch, tap, and keyboard all use the same rules.
Reference Ranger and Skeleton portraits are cropped from the user-supplied Spell Valley gameplay screenshot and appear only in this unofficial browser mockup—not in the downloadable concept pack.
What Connor gets today
The pitch, the proof, and the source files
The primary handoff is tile-first. The Blender files remain useful as a character-design source and bonus production artifact.
ZIPComplete tile pack
LIVEPlayable rules prototype
READMERules + import notes
BLENDEditable 3D source
GLBCompact 3D export
JSONVerification report
Real-size checkSame art. Smaller decisions.
The delivered alpha crops are shown at their actual pixel dimensions. The prototype circle can be replaced with Connor’s real UI mask.
Fastest useful test with Connor
Drop the PNG into the real mask, then graybox these two scenarios.
- Check the 64–96 px portrait against the live roster.
- Confirm action cost, move range, and turn terminology.
- Decide whether stepping on the marker is the right enemy answer.
Spell Valley today · reviewed July 20, 2026
Grounded in the public game, explicit about the gaps
Official store materials show portrait-driven hero tiles on a hex-shaped tactical map, fast turn-based battles, and a roster of 35+ heroes. The demo now matches that interaction grammar with contained unit cards and correct hex spacing; two small reference portraits come from the gameplay screenshot supplied for this review.
Publicly observableDesign responseValidate with Connor
Portrait tiles on a hex board
Visible in official store screenshots and the supplied gameplay references.
Contained cards + true axial spacing
Every unit stays inside one discrete cell; resize-aware geometry preserves the hex grid.
Exact board rules
Move range, blockers, objectives, and real UI safe areas.
Short turn-based battles
The listings pitch roughly five-minute matches.
One visible trigger
Place, contest, consume. No hidden proc or nested status stack.
Action economy
Cost, cooldown, line of sight, and turn boundary.
35+ hero roster
Distinct reads matter in a crowded collection.
A route-support silhouette
Antler crown, bright eyes, and waylight remain legible at 64 px.
Roster overlap
Whether a live hero already owns this tactical niche.
Current Dreadmire theme
Recent public release notes foreground darker swamp characters.
A hopeful counterpoint
Mint waylight and old-road lore offer a different visual temperature.
Roadmap fit
Theme cadence and any unreleased Grove plans.
Evidence key: public statements above come from the official Google Play listing and official App Store listing. The Ranger and Skeleton reference portraits were cropped from the user-supplied gameplay screenshot for this mockup only. The character, role, demo rules, and board composition are proposals.
Open the full Waylight v0.2 test specification
- Placement
- One adjacent, empty, connected walkable hex; not an objective, hazard, tower, or castle.
- Duration
- Expires at the start of Glimmerroot’s next activation.
- Ally trigger
- The first other ally entering by normal movement consumes it, then moves one extra connected hex.
- Enemy answer
- Enter first by normal movement to extinguish it and gain no bonus.
- Guardrails
- Non-blocking; one active; no stacking, chaining, teleport, spawn, or forced-move triggers.
- Tuning
- Ability cost, cooldown, line of sight, and whether the bonus ends movement remain deliberately unresolved.
Bonus production source
The 3D model still earns its keep
Blender provides an orbitable character bible, repeatable poses, and verified interchange files. It is the design source—not the proposed battlefield format.
Loading interactive source…Drag to orbit · scroll to zoom
Asset flow
- 013D character master
Locks silhouette, materials, pose, and lighting.
- 022D portrait master
Preserves the crown, core, staff, and waylight seed.
- 03Tile-safe crop
Face and chest remain identifiable down to 64 px.
- 04Playable browser proof
The same motif becomes a contested marker on a real hex graph.
Source mesh1,157 verts / 2,054 tris
Clean import2,134 decoded tris
Rig / clips9 bones / 3 clips
Dependencies0 texture files
What the AI workflow actually did
The correction became part of the result
This factual audit trail excludes credentials, private system instructions, hidden reasoning, and raw server logs.
- 01
Reviewed the brief with a subagent
The independent passes set quality bars for originality, real assets, game fit, legal-move clarity, enemy counterplay, accessibility, and VPS isolation.
- 02
Researched the live public context
Reviewed official store copy, release notes, and screenshots. Public evidence now sits after the playable and downloadable work instead of interrupting the pitch.
- 03
Built and iterated the Blender source headlessly
Used the normal Blender 5.2 LTS application in background mode to create geometry, materials, a 9-bone rig, three animations, renders, BLEND, GLB, and FBX files.
- 04
Pivoted from 3D-first to tile-first
Created a painterly portrait from the project’s own Blender reference, refined its small-size focal hierarchy, and exported five masked tile sizes plus a Waylight marker.
- 05
Specified the mechanic and its counterplay
Defined placement, duration, ally trigger, enemy denial, expiry, stacking, and edge-case guardrails while leaving private balance variables explicit.
- 06
Rebuilt the demo as a playable hex tutorial
Added a 19-cell axial hex model, computed legal targets, drag/touch/tap/keyboard input, invalid-move feedback, a reach meter, ally success, enemy denial, replay, and complete scenario resets.
- 07
Corrected the board against supplied gameplay
Replaced compressed percentage spacing with resize-aware flat-top axial coordinates, contained each portrait inside a unit card, added real-game reference portraits, and moved the next-action cue directly above the board.
- 08
Verified and deployed the isolated page
Clean-imported both 3D formats, checked the package and browser interactions, then updated only this VPS subdirectory without changing the existing homepage or server configuration.
Prompt record
All user-authored prompts for this project
Five project prompts are reproduced verbatim below. Platform metadata, system/developer instructions, secrets, and private reasoning are intentionally not published.
I have a friend Connor who works on the mobile game spell valley.
Can you download blender make a 3d model of a new character he could add to the game then make a new page on my hetzner VPS then send the link to me?
On the page you should note all of the prompts i gave you and what work you did in a clear way.
Ideally, this is the only prompt needed so when in doubt dispatch a subagent to review the original task and what sticking points you have and what might get you unstuck
This is great, now let's assume that this is the first version of me helping someone onboard to using ai productivity and we really want to wow them both with the 3d model and with the work done. Do another review pass with a critical eye and clean this up note details about how this relates to spell valley as it is today and what else we might be able to do to improve this. Ideally come back with something even better with an even more impressive wow factor. I would probably vote for iteration on this rather than making something new. As an aside, I am surprised to have not watched you do this work in blender. Are you working in a headless version or something? Don't change anything, I'm just curious.
Oh and lol spell valley doesn't have 3d models of characters they have tiles. So we can keep the character model here, but Ideally we instead of artwork on a tile and walk out the mechanic.
We should make this playable in the browser rather than clicking through the tabs. Make it so you can only do the things suggested with queues to the user so they understand, and maybe leave the tabs to reset things, but I wanted to drag and drop things as you do in the actual game in the demo. This also makes the impact of the enemy less clear because I can't test it. Also the board if you look at it is a hexagonal board. We should leverage that in this demo so it really looks and feels like the actual game.
So in summary. 1. Redo the demo portion so it feels playable. 2. Re-order the pieces so it shows the art and useful bit up front, followed by the work delivered, then go into details that aren't relevant etc later. So we should start with attention grabbing executive summary and flashy marketing work and then go deeper into still explicitly useful details as we go.
I would put the art work and mechanic demo above **Grounded in the public game, explicit about the gaps etc**
this is what the game looks like why are the tiles overlapping? It isn't clear what you are supposed to do at any stage of the demo. The attacker or enemy should be tiles from the game ideally. Feel free to search online to pull those assets.