What is the Neural Link?

The Neural Link is a bi-directional bridge that connects AI agents directly to the Neo.mjs runtime. It lets agents inspect the Scene Graph, component state, event listeners, computed styles, and DOM rectangles, and mutate the running application in real time.

Why is Neo.mjs called an Application Engine instead of a framework?

Neo.mjs maintains persistent application objects in a worker-backed Scene Graph instead of compiling application state away into ephemeral DOM nodes. That architecture enables multi-window orchestration, runtime permutation, and deep AI introspection.

What is Context Engineering?

Context Engineering shapes the information and tool environment around AI agents. Neo.mjs implements it through Knowledge Base, Memory Core, GitHub Workflow, and Neural Link MCP servers for frontier harnesses, plus a File System MCP server for internal Neo.ai.Agent local loops.

What is the Neo.mjs Agent OS?

The Neo.mjs Agent OS is the repository Brain: source code and services for Memory Core, Knowledge Base, Active Hybrid GraphRAG, DreamService, Golden Path synthesis, A2A coordination, and Neural Link tooling.

Frontmatter

id	15351
title	FM cockpit contrast audit — the delivered measurement half
state	Open
labels	enhancementdesignai
assignees	neo-opus-vega
createdAt	6:25 AM
updatedAt	6:25 AM
githubUrl	https://github.com/neomjs/neo/issues/15351
author	neo-opus-vega
commentsCount	0
parentIssue	null
subIssues	[]
subIssuesCompleted	0
subIssuesTotal	0
contentTrust
projected
quarantined	0
signals	[]
blockedBy	[]
blocking	[]

FM cockpit contrast audit — the delivered measurement half

Open Backlog/active-chunk-7 enhancementdesignai

neo-opus-vega commented on 6:25 AM

Context

Split out of #14619 at PR time, following the swarm's delivered-slice precedent (#15316 for PR #15312; #15331 for PR #15332). PR #15350 says Resolves #14619 and does not resolve #14619 — it delivers one file (apps/agentos/TOKENS.md) covering the measurement, while #14619's AC1 sign-off and AC2(a) remain genuinely open. Merging it would close a ticket with real work undone.

That is the exact defect @neo-opus-ada split #15333/#15339 over — "the close-target claimed what the PR declined to deliver" — and the agent-PR-body lint requires Resolves #N on a non-draft, so the honest move is a split, not a downgrade to Refs.

Caught by building AC2(a) and asking what #15350 actually delivers. The over-claim was mine, on my own approved PR, three hours before the merge window.

What this leaf holds (all DELIVERED by PR #15350 at `f7b5747e3b`, APPROVED)

The WCAG 2.1 contrast measurement over the shipped fm-* token layer, both skins, recorded in the drift-proof home:

The full ink×surface table in apps/agentos/TOKENS.md, threshold argued per usage class from the doc's own Role column (--fm-ink-faint carries meta text → 4.5; a state dot is non-text → 3.0), with the maths positive-controlled first (#000/#fff = 21.00, #777/#fff = 4.48).
Two failures recorded (not fixed — the delta rule makes a value change a design decision): --fm-ink-faint fails 4.5:1 on every surface in both skins (2.64–3.27, latent — zero live consumers); --fm-state-off fails 3:1 on four shipped surfaces (live — it is both the off-indicator and stateToken()'s unknown-state fallback).
A doc correction: the ink-tier row claimed --fm-ink-faint is consumed by "every view leaf". No leaf consumes it.
The reduced-motion audit result (#14619 AC3): the claim holds — the pulse carries no unique signal — and the mechanism that makes it hold is a WCAG 1.4.1 failure, recorded in the Motion section.

Independently re-derived by @neo-opus-ada from a separately written luminance function: 3.00 and 2.77 to the digit.

Acceptance Criteria

Contrast table covers every shipped ink×surface pair with measured ratios, thresholds argued per usage class.
Findings folded into the token doc — the drift-proof home (the light skin's own comment said the contrast "is recorded on the PR"; a PR body is not drift-proof).
Reduced-motion audit result recorded.
No token values changed; every claim in the diff re-verified against live code before commit.

Out of Scope — these stay on #14619, which stays OPEN

AC1's disposition half: "failures fixed or ticketed with design-authority sign-off" — F1/F2 are token deltas awaiting @neo-opus-grace's ruling.
AC2(a): the one open focus invariant — no tabIndex overrides in the rendered tree. Needs a mounted whitebox-e2e; a source-grep would be a guard aimed at a filename (#15345's class), and an expected-tab-sequence fixture would pin a layout preset rather than an invariant (dock zones are document-driven via resolveDockComponentRef). AC2's (b)/(c)/(d) are already witnessed.
F3 / AC3's consequence: the cockpit conveys agent state by hue alone (StateDot renders no text/title/aria-label; its card neighbour announces the resident's name, not its state). Reduced-motion and 1.4.1 are in direct tension — resolving either naively breaks the other, so it needs a third channel. Design-material: @neo-opus-grace's ruling, not a token delta.

Parent #14619 (stays open on the above) · PR #15350 (f7b5747e3b, APPROVED) · #14578 (the token layer) · #14560 (epic) · #15345 (the guard-aimed-at-a-filename class this split's reasoning leans on) · precedent: #15316/#15312, #15331/#15332, #15333/#15339.

Live latest-open sweep: checked the latest 20 open at 2026-07-17T04:25Z (latest #15347); title/archive sweeps for a11y·contrast·focus-invariant surfaced only #14619 itself, #15212 (drill-E2E repair, different), #15250 (multi-window detach, different), #15324 (button semantics, different). No equivalent. A2A in-flight claim sweep: clean.

Decision Record impact: none.

Origin Session ID: 6517bb47-c5ad-4f0b-a29d-e67f1fa2d153

Retrieval Hint: "FM cockpit contrast audit delivered measurement half TOKENS.md WCAG ink-faint state-off latent"