Current Recheck Note — 2026-06-08

Full report

On this page

This report is retained as noindex historical research with stale-recheck-required handling. The current UAIX AI Memory Package Wizard public page describes an eight-step local-first package planning flow. Any older six-step wizard language in this report is superseded by the current public support boundary and must not steer curated /docs or public copy without revalidation.

---

UAIX v1.0 Beta Specification and Wizard Update Assessment

Executive summary

UAIX.org already presents a substantial public v1.0 surface: the current release is UAI-1, the homepage describes UAIX as the public evidence and handoff layer for agentic systems, the machine catalog/discovery routes publish six active schemas/profiles, and the AI Memory Package Wizard is a six-step local package-planning tool with review/export, draft restore, optional Agent File Handoff planning, and optional LLM Wiki planning. The public site also makes a strong support-boundary claim: no hosted import, no automatic repository writes, no automatic LLM Wiki sync, no SDK/CLI/certification/endorsement claims.

The most important finding is that the current public site and the uploaded draft-spec materials are materially out of sync. The public Project Handoff package model is centered on AGENTS.md, readme.human, and .uai/... files such as .uai/context.uai, .uai/constraints.uai, and .uai/memory.uai, while the uploaded draft documents introduce a newer handoff architecture built around root-level totem.uai and taboo.uai, a startup state machine, an optional export manifest, and validator/wizard behaviors not yet reflected on the public site or in the public wizard outputs. Similarly, the public UAI-1 spec publishes six profiles and does not currently expose public support for idempotency_key, correlation_id, or dedicated discovery/delegation profiles, while the uploaded orchestration draft proposes those additions.

For a v1.0 beta release, the highest-confidence path is not to replace the current public architecture. Instead, UAIX should ship an additive beta that preserves the current six-profile UAI-1 surface and the current AI Memory / Project Handoff package structure, while introducing draft concepts in a compatible way. Concretely: keep the existing wizard shape and existing package files; add totem.uai and taboo.uai as optional-but-recommended beta outputs; add orchestration semantics as optional fields and conventions inside the existing envelope rather than brand-new mandatory profile families; tighten copy where current pages overstate parity that the roadmap still marks as “next”; and publish all changes through the canonical page + machine-artifact + validator/fixture + changelog path that UAIX already uses for support claims.

A second important finding is governance-related: there is no public repository or public issue tracker published today. UAIX’s own References and Contributors page says contributor intake is documentation-led and that the public handoff is the canonical site, machine artifacts, implementation tracks, and release notes rather than a publicly published repo/issue queue. That means this research could prioritize the official site and official machine routes, but could not inventory a public repo or issue tracker because the public site explicitly says those surfaces are not yet published.

The release recommendation is therefore straightforward: treat the public site as the v1.0 source of truth, treat the uploaded drafts as the product-owner target state, and ship a three-phase beta that first aligns terminology and public copy, then adds compatible beta outputs and readiness checks to the wizard, and only after that publishes any new machine-readable UAI-1 fields or schemas. That sequencing minimizes breakage, preserves the current boundary claims, and gives beta users a clear migration story.

Source basis and current public baseline

The reviewed primary sources were the canonical UAIX public pages, the UAIX machine-facing JSON routes, and the three uploaded internal draft documents. The strongest public sources for this assessment were the homepage, AI Memory page, Project Handoff page, Agent File Handoff page, AGENTS.md spec, UAI-1 spec, API Reference, Roadmap, Changelog, References and Contributors, Privacy and Data, and Analytics pages, plus the /wp-json/uaix/v1/catalog, /wp-json/uaix/v1/discovery, and /wp-json/uaix/v1/roadmap routes. The uploaded draft-spec sources were the handoff/Totem-and-Taboo draft, the memory-package blueprint/state-machine draft, and the multi-agent orchestration/safety draft.

The current public baseline is consistent on several points. UAIX presents UAI-1 / version 1.0 as the current public release, with six schemas/profiles, two transport bindings, five trust channels, four conformance levels, and eleven error codes published via the machine catalog/discovery surfaces. The homepage also publishes a local build record of UAIX-2026.04.29-predeploy, while the discovery route reports a generated timestamp of 2026-05-06T15:05:57+00:00.

The public support boundary is equally consistent. Across the wizard page, roadmap, changelog, privacy/data, and analytics governance pages, UAIX repeatedly states that its current support is a public standards/documentation/evidence surface, not a hosted importer, not an automatic repository writer, not an automatic LLM Wiki sync tool, not an SDK/CLI program, not a certification/endorsement body, and not a replacement for agent runtimes or adjacent protocols such as MCP, A2A, and OpenAPI.

The public UAI-1 contract is still relatively compact. The canonical page publishes six current profiles — uai.intent.request.v1, uai.intent.response.v1, uai.capability.statement.v1, uai.error.v1, uai.conformance.result.v1, and uai.task.status.v1 — and describes the public record as centered on explicit identity, workflow continuity, trust posture, provenance, integrity, typed errors, and validator-backed evidence. The validator page reinforces that current validation checks cover schema, field order, trace context, delivery expiry, capability-declared transport bindings, conformance levels, and trust-policy hints.

The public AI Memory / Project Handoff baseline is also clear. The AI Memory page positions AI Memory as compact, portable, file-based durable context, with Project Handoff as a subtype. The current Project Handoff starter package on the public page includes root Markdown files plus .uai/context.uai, .uai/constraints.uai, .uai/memory.uai, and package-specific .uai files such as .uai/stack.uai, .uai/architecture.uai, .uai/progress.uai, .uai/operations.uai, .uai/test-plan.uai, .uai/style.uai, and .uai/decisions.uai. Agent File Handoff is a separate spec with agent-file-handoff/Content/, Improvement/, Archive/, plus .uai/file-handoff.uai and .uai/intake-index.uai.

The largest availability limitation is this: a few routes that the public site names as canonical — especially .well-known and some governance pages such as Launch Readiness/Accessibility — could not be directly fetched by this research tool even though other UAIX sources list them as public routes. That is a verification gap, not confirmed site failure, and should be treated as unknown pending direct browser/host validation.

Wizard inventory, UI text, and published feature-state inventory

The current wizard is a six-step guided builder for supported UAI AI Memory starter bundles. It has a human UI and a page-embedded visitor-AI digest (script[data-ai-digest]), a base route plus fragment routes for preselecting Agent File Handoff and/or LLM Wiki, local browser draft restore, page-level validation, local-only generated artifacts, canonical starter ZIP downloads, and no hosted import/write behavior. That public behavior is described both on the wizard page and in the changelog entries from 2026-04-30 and 2026-05-01.

No public internal feature-flag registry was found in the reviewed official sources. UAIX explicitly says that the public contributor/review surface is the published site plus machine artifacts and release trail, not a public repository or issue queue. Accordingly, the table below inventories published feature-like states and observable switches, not hidden internal flags.

Discrepancy matrix

The uploaded draft materials represent a coherent next-stage design direction, but they do not match the current public site in several important ways. The matrix below maps the highest-priority mismatches between the current public implementation and the internal target state.

Proposed v1.0 beta spec updates and wizard guidance

The safest product/spec strategy is to keep the current public v1.0 contract intact, then layer in the draft concepts as additive beta guidance. That means: preserve the current six-profile message family, preserve the current Project Handoff package shape, preserve the current wizard route/step architecture, and introduce safety anchors and orchestration semantics in ways that do not invalidate current public packages or the current validator surface. This recommendation directly follows from the public support-boundary language, the current published package/file model, and the roadmap’s own “current / next / planned” distinctions.

flowchart TD
    A[Wizard entry] --> B[Preset]
    B --> C[Basics]
    C --> D[Operations]
    D --> E[Protocols]
    E --> F[Receiver]
    F --> G[Review and Export]

    D --> D1[Optional beta safety anchors block]
    D1 --> D2[Generate totem.uai]
    D1 --> D3[Generate taboo.uai]

    F --> F1[Optional Agent File Handoff plan]
    F --> F2[Optional LLM Wiki plan]

    G --> G1[Local preview]
    G --> G2[Copy-paste deck]
    G --> G3[Canonical ZIP]
    G --> G4[Package model and overlay]
    G --> G5[Readiness checks]

flowchart TD
    A[AGENTS.md] --> B[readme.human]
    B --> C{totem.uai present?}
    C -- Yes --> D[Load totem.uai]
    C -- No --> E[Load current .uai context files]
    D --> F{taboo.uai present?}
    F -- Yes --> G[Load taboo.uai]
    F -- No --> E
    G --> E[Load .uai context,\nconstraints, memory,\noperations, test-plan]
    E --> H[Load startup packet / receiver brief]
    H --> I[Do targeted checks]
    I --> J{taboo violated?}
    J -- Yes --> K[No-op and escalate]
    J -- No --> L[Proceed with reviewed work]

The exact wording changes below are written as proposed normative/public-copy inserts for a v1.0 beta release.

The wizard copy should be updated in a very targeted way so the user-facing flow remains familiar. The table below keeps the six-step architecture and adds only the minimum new language needed for a v1.0 beta.

Two implementation details should remain explicit in the spec and wizard guidance. First, do not publish new canonical public profile families yet unless schemas, registry entries, examples, API docs, and validator expectations all land together. Second, do not make totem.uai / taboo.uai mandatory in v1.0 beta because that would invalidate the current public Project Handoff bundle family and the current wizard-generated file set. The roadmap and changelog both argue for additive, evidence-backed promotion rather than premature support-language widening.

Rollout plan, changelog, testing, and migration notes

A practical beta rollout can fit into a short three-week schedule while remaining faithful to UAIX’s existing evidence discipline. The recommended sequence is: first publish copy/boundary alignment, then ship wizard artifacts and review checks, then publish any associated machine-readable additions. That order mirrors the roadmap’s promotion gates and minimizes the risk of the public pages claiming support ahead of proof.

A suggested beta-user changelog set for publication is below.

A concise beta-user rollout note should read like this:

UAIX v1.0 beta keeps current UAI-1 and Project Handoff packages valid. Existing starter ZIPs, current .uai files, AGENTS.md, and readme.human remain supported. New beta artifacts such as totem.uai, taboo.uai, and additive orchestration fields are optional unless a specific implementation lane says otherwise. Do not treat optional beta artifacts as public conformance requirements until the matching schemas, examples, validator checks, and changelog entries are published.

The QA/testing checklist should cover both public-site behavior and generated-artifact correctness.

The migration story should be simple and conservative.

Effort, priorities, risk assessment, and open questions

The change set can be ranked cleanly by value and implementation risk. The items below assume the goal is a credible v1.0 beta with minimal public breakage.

The primary risks are all manageable. The biggest risk is support-boundary drift: if the beta introduces safety anchors or additive orchestration fields but public pages start implying hosted enforcement, imports, or runtime ownership, UAIX will contradict its own roadmap and governance posture. The second risk is serializer/documentation drift: the public pages currently describe strong synchrony across wizard, digest, and ZIP surfaces, while the roadmap still treats parity proof as incomplete. The third risk is breaking existing handoff packages by making draft concepts mandatory too early. The fourth risk is verification blind spots around some named discovery/governance routes that could not be directly fetched during this research session.

The open questions are short and important. No public repo or issue tracker is published, so release/change management must continue to flow through the canonical site, machine artifacts, and changelog unless UAIX chooses to publish a public source/review surface later. The exact embedded script[data-ai-digest] payload on the wizard page was described publicly but not directly extractable through this research tool, so its exact keys and parity with the visible wizard remain partially unknown. The site also publicly names .well-known, Launch Readiness, Accessibility, and other routes that this tool could not directly verify, so final beta QA should include a manual/browser-level production-route check. Finally, the internal draft documents are directionally strong, but their mandatory language would be breaking if applied verbatim to the current public v1.0 package model; the recommended beta path therefore treats them as target-state guidance, not immediate public replacement text.