Wildz

A living game world built on proof-native commerce.

Wildz is a real, full-screen game PWA and the first end-to-end reference product built by taking the open-source Receiz Commerce Kit beyond a storefront. It turns the same identity, proof, ownership, publication, settlement, MCP, and AI-skill primitives into a persistent world where players explore, collect, compete, share, and trade.

Wildz is both a product and a reference implementation. The game is the product; the repository is the evidence that the Receiz application model can be reshaped into something genuinely different without replacing its proof authority.

Why this repository exists

Most SDK examples stop at a code snippet or a sample checkout. Wildz is deliberately larger: a production-shaped application with identity recovery, portable artifacts, deterministic gameplay, multiplayer, public projections, offline reads, commerce coordination, release gates, and agent operating doctrine.

It demonstrates four layers working together:

Layer	What Wildz uses it for	Authority boundary
Receiz Commerce Kit	The forkable application kernel and original commerce architecture	Starting point, not a runtime dependency
@receiz/sdk@127.0.0	Identity, proof objects, durable subjects, authority sessions, durable execution, recipient resolution, and executable Phi value rails	The typed application/runtime boundary
@receiz/mcp-server@127.0.0	Capability inspection and authorized operator workflows	Tooling only; it cannot manufacture proof or authority
@receiz/ai-skills@127.0.0 plus the checked-in Wildz skills	Proof-aware build, market, and release procedures for coding agents	Operating guidance only; verification still wins

All three Receiz packages resolve at exact version 127.0.0 from the public npm registry, with their published SHA-512 integrity values pinned by pnpm-lock.yaml. This release adopts the v127 ruleset, whose registry digest is 8d0b5b839d02d9efbd4306cc99410595a183705c2670b76d2567eaaaade99065 and operation-matrix digest is eadd171a45fcc51e275a1c57de1eb8e67614757a5723d141793641edf7207a10. The application contract in receiz.app.json selects artifact-first authority, declares the complete 60-operation V127 matrix, and explicitly disables database authority. See the V124 runtime composition for qualified Twin enrichment, proof memory, authority sessions, durable execution, replay/private additions, and MCP custody boundaries.

V123 keeps that source-first continuity and adds exact proof-authority exchange, scope introspection, namespace resolution, Settlement/Reserve execution, and idempotent value-execution recovery to the retained V122 planning rails. Wildz uses only the exact SDK-custodied primitives that the published client exposes; missing durable application dependencies remain fail-closed instead of being recreated locally.

What was built

Wildz replaces the source repository's commerce-site experience with a game-first product while preserving and extending its Receiz rails:

• A no-signup first landing with local player genesis and later Receiz identity activation.
• Identity Seal/key continuation and verified Vault recovery.
• Living, portable creature cards with lineage, mastery, progression, export, import, and cross-application continuity.
• Universal playable proof claims for exact card custody today and SDK-issued Phi, resource, access, and world-right transition sets through the same native Claim surface.
• A proof-grounded creature Twin: talk to the exact being from its card, hear its deterministic neural voice, and see genome/state-shaped expression synchronized to speech.
• Search-ready canonical metadata, game structured data, standard and image sitemaps, and bespoke Open Graph/iMessage/Discover artwork that remain outside the gameplay hot path.
• A deterministic world with ecology, settlements, social memory, routes, raids, bosses, and canonical event projections.
• Mortal Arena, Hearttree, party play, competition, and multiplayer commands.
• Sanitized public player profiles and public card routes.
• A social market overlay with listings, offers, trades, checkout, and settlement contracts.
• An installable PWA with explicit install/update consent and a narrow, tested offline-read boundary.
• A constitutional Receiz application contract, repository checker, secret scanner, conformance tests, and one-command release gate.
• Three repository-native AI skills that teach agents how to build, operate the market, and qualify a release without outranking verified proof.

This is not a reskinned storefront. It is an architectural fork: the inherited kernel has been exercised against a different interaction model, state model, threat model, and product loop.

Quick start

Requirements: Node.js 20.19.0 or newer and pnpm 10.29.1.

git clone https://github.com/kojibai/wildz.quest.git
cd wildz.quest
corepack enable
pnpm install --frozen-lockfile
cp .env.example .env.local
pnpm dev

Open http://localhost:3000. The local product can boot without production credentials. Shared-world publication, live identity coordination, market mutations, and settlement remain unavailable until their server-side Receiz capabilities are configured.

Before changing code, establish a clean baseline:

pnpm release:check

That command runs the Node test suite, typecheck, Receiz v127 contract checker, MCP conformance, lint, tracked/untracked text secret scan, production build, and default Receiz doctor.

The system in one view

flowchart LR
    P["Player or operator"] --> UI["Wildz Next.js PWA"]
    UI --> G["Deterministic game systems"]
    UI --> A["Wildz Receiz adapters"]
    A --> SDK["Receiz SDK"]
    SDK --> O["Verified proof objects"]
    SDK --> R["Configured Receiz rails"]
    MCP["Receiz MCP"] --> A
    S["Wildz + Receiz AI skills"] --> MCP
    S --> UI
    O -->|"authority"| A
    R -->|"admitted shared state"| A
    A -->|"verified projections"| UI

Loading

The important rule is directional: AI skills and MCP can help inspect, prepare, and invoke work, but neither can promote model output, local storage, a receipt, or a queued proposal into authority. Complete artifacts are verified before their payloads are interpreted. Canonical transitions require current evidence and fail closed when a capability is absent.

Read the detailed architecture, Receiz rail map, and MCP contract.

Creature-compatible experiences

Start with the developer guide, the creature entry point, and the test-covered trail example. They cover verified identity, current capabilities, offline play, and the boundary between a local experience and portable authoritative progress. For card URLs, use the public-card publication guide and public recovery example: a valid card proof is distinct from permission to publish it.

SDK, MCP, and AI skills

SDK application boundary

Application-facing Receiz code lives in src/lib/receiz. UI and game modules consume these adapters instead of scattering SDK calls across components. New v121 artifacts use the native Record → Seal flow, preserve SDK-returned bytes exactly, and are independently reopened before acceptance.

The checked-in contract and generated evidence bind the application to the v127 ruleset, registry digest, 60-operation matrix, protocol limits, retained numbered artifact laws, living-subject authority, profile/economy showcases, and native-capture/PBI-authorship rules. Twin output is non-authoritative; factual memory cites admitted events; multi-subject effects are atomic; and bearer transfer preserves subject identity while revoking the former owner.

pnpm receiz:check
pnpm receiz:doctor

MCP operator surface

Run the pinned MCP server from an MCP-capable agent host:

pnpm exec receiz-mcp

Public reads do not require a bearer token. Authorized delegated operations require a scoped Receiz credential supplied to the MCP process. MCP packages must stay out of browser and application bundles. See docs/MCP.md for the nine artifact tools, 37 living-subject tools, and credential boundary.

AI-native repository operations

The ai-skills directory contains installable, versioned procedures for agents working in this codebase:

• wildz-builder-skill preserves custody, continuity, append-only history, and deterministic projections.
• wildz-market-operator-skill governs bearer ownership and capability-gated market work.
• wildz-release-skill enforces package, digest, artifact-law, conformance, verification, and release parity.

These skills make the repository teachable to an agent without turning the agent into a source of truth. See the AI skills guide.

Local, live, and offline boundaries

Wildz adds no external application database. Owner-scoped continuity is retained in browser IndexedDB; durable public, world, social, and economy state depends on configured Receiz rails.

Surface	Local development	Configured live environment	Offline
App shell and local verified owner state	Yes	Yes	Previously installed state only
Previously visited public profiles/cards	Yes	Yes	Readable from the allowlisted cache
Identity/Vault verification	Fixture and local flows	Yes	Artifact-dependent; no remote continuation
Live world and social presence	No durable claim	Yes	No
Publication and public projection writes	Fail closed	Capability required	No
Listing, trade, transfer, payment, settlement	Fail closed	Every required capability and proof required	No

V121 exposes direct bearer transfer preview, instrument issue, inspection, claim, cancellation, and status. Wildz uses those surfaces only with genuine scoped capabilities and still independently verifies the complete returned artifact. Marketplace listing/payment settlement remains unavailable wherever its separate conditional market append is absent; the app never substitutes IndexedDB, process memory, a bearer receipt, or checkout success for settlement authority.

Repository map

Path	Purpose
app	Next.js pages, public routes, and 34 API route handlers
src/features	Identity, world, games, market, profile, shell, and PWA product modules
src/lib/receiz	Receiz adapters, artifact custody, sessions, repositories, verification, and publication
tests	272 contract, game, continuity, PWA, market, and release test files
ai-skills	Agent-readable build, market, and release doctrine
docs	Architecture, Receiz boundaries, interoperability, and release evidence
scripts	Release gate, doctor, checker, conformance support, and secret scanning

Release status

v10.1.0 — Continuity and Control advances the v10 Living World with faster world work, distinct Glide and Flight controls, stronger identity and Vault saves, canonical offline recovery, and wallet fixes. Application and installed-PWA versions are 10.1.0 and v10.1.0-r1; the coordinated Receiz toolchain remains exact 127.0.0.

Read the v10.1.0 release notes and complete post-v10.0.0 commit ledger, v10.0.0 release recap, v127 integration evidence, game-image verification boundaries, and changelog.

Build your own Receiz-native product

Use this repository when you want to study a complete vertical product rather than a minimal SDK call. A disciplined fork should:

1. Replace the Wildz product domain while keeping Receiz calls behind a narrow adapter.
2. Declare features and authority in receiz.app.json.
3. Treat carried proof as untrusted until the SDK verifies the complete artifact.
4. Keep MCP and agent actions below the same capability and confirmation gates as human actions.
5. Add domain tests before expanding a canonical mutation surface.
6. Make pnpm release:check a required branch and release gate.

The architecture guide identifies the extension seams and non-negotiable authority rules.

Contributing and support

Wildz is open for rigorous contributions. Start with CONTRIBUTING.md, use SUPPORT.md for help, and follow SECURITY.md for private vulnerability reporting. Participation is governed by CODE_OF_CONDUCT.md.

Provenance and license

Wildz includes code derived from kojibai/Receiz-commerce at commit fb366506e218d82ecac20c60bc74c5977627713e. The fork relationship, retained license obligations, and Wildz-specific composition are recorded in NOTICE.md.

Released under the MIT License.