Reactor Dsl

What this skill carries

The full alphabetized signatures index lives at:

references/reactor.api.txt    (~70K tokens, ~6.3K lines)

It covers every public factory, modifier, hook, theme token and enum, plus a

## Public types section carrying the constructors, properties, methods and

events of every other public type — including public static classes, so

process-wide entry points such as ControlRegistry.Register and

ReactorApp.Run are listed there too.

It is large: grep it for the symbol you need rather than reading it end to end.

It is the source of truth for the public API surface, regenerated from Reactor.dll by mur --regen-api.

In addition, the skill carries focused topical references for surfaces

big enough to need their own page:

| Topic | File | When to load |

|---|---|---|

| Keyed list reconciliation | references/keyed-lists.md | Choosing between IReactorKeyed, an explicit keySelector, and the three .WithKey(...) overloads; debugging a list that re-mounts on insert; understanding the bulk-replace bailout or the duplicate-key diagnostic. (Spec 042.) |

| Element refs and reference props | references/element-refs.md | Wiring UseElementRef<T>(), .Ref(...), .Target(...), accessibility relationships, XYFocus*, or custom descriptor .Reference / .ReferenceList entries. (Spec 057.) |

When to load this skill

You probably don't need to. The reactor-getting-started skill carries the hooks table, the most-used factory signatures, the React-to-Reactor mapping, modifiers, theme tokens, and the critical gotchas — that's the 90% case.

Load this skill only when:

• You've checked reactor-getting-started and the factory/modifier you need isn't listed.
• You see a REACTOR_* analyzer ID and want to confirm the exact API surface in question.
• You're using a less-common control (DataGrid column overrides, AnnotatedScrollBar, AcrylicBrush parameters, etc.).

How to use the api index

Read the file once, scan for what you need, then keep working from memory. Do not re-page through it — the file is large and re-reading injects ~12K tokens per call. Per-pattern lookups are far cheaper than full re-reads.

If you only need to confirm whether a single name exists, use a grep for the symbol against the file rather than viewing it whole.

Fluent naming convention (callbacks)

Callback properties are named OnX on the record (e.g. OnClick, OnChanged,

OnSelectedTagChanged). The matching fluent extension drops the leading On

and is named after the event:

// Record property is OnClick — the fluent is .Click(...)
Button("Save").Click(() => Save());

// Record property is OnSelectedTagChanged — the fluent is .SelectedTagChanged(...)
NavigationView(...).SelectedTagChanged(tag => Navigate(tag));

This is intentional — C# doesn't allow an extension method to share a name

with an instance property, so the On prefix lives on the property and the

event-style name lives on the fluent. The two surfaces are equivalent;

prefer the fluent for new code. Passing null to any OnX fluent clears

the handler.

Common naming gotchas

• FlexElement record properties (set via with { ... }):

Direction, JustifyContent, AlignItems, AlignContent, Wrap, ColumnGap, RowGap

⚠️ It's JustifyContent — NOT Justify.

Example: FlexRow(a, b, c) with { JustifyContent = FlexJustify.SpaceBetween, ColumnGap = 8 }