VALKYRIE

Automation + Intent Routing

The shell should learn your patterns, not make you repeat them forever.

Valkyrie automation is about removing ritual. If the browser should maximize, do that. If a download should open in the right place, route it. If a setup matters, restore it. Intents, rules, and explainability make the shell programmable without turning it into a black box.

Daily Value

Open the browser, get the right layout.

If a window should maximize, move workspaces, or trigger a scene, Valkyrie can just do that.

Route actions by meaning.

Open a URL, video, document, or handoff and send it where it belongs instead of leaving behavior to chance.

Know why something happened.

Rules and routing are explainable. Valkyrie can tell you which rule fired, what won, and what lost.

Intent Layer

Intent Sources

Fileman, phone actions, clipboard, Glimpse, remote flows, and explicit CLI or IPC sends can all feed the same routing model.

Intent Router

Rules can consider payload shape, extension, domain, source, and context so actions land where they make sense.

Targets

RefugeWeb, media tools, Waypad, Gather, notes, command routes, and other desktop surfaces can all become policy targets.

Rules

Map-time behavior

When an app appears, Valkyrie can maximize it, place it, move it to a workspace, or trigger follow-up behavior.

Focus, timers, profiles

Automation is not limited to app launch. It can react to focus shifts, timed events, or broader workflow/profile changes.

Repeatable by design

The point is not novelty. The point is that useful desktop behavior becomes repeatable and teachable.

Explainability

Automation only stays trustworthy if it is inspectable. Valkyrie keeps rule decisions, routing outcomes, and context visible so “what just happened?” can become “I know exactly why that happened.”

SNAPSHOT_CREATE name=pre_change result=ok
RULE_MATCH name=browser_to_workspace_three result=matched
INTENT_ROUTE name=open.file decision=mpv reason=header+policy
RULE_SKIP name=browser_to_workspace_three reason=focus_mismatch

AI as a Layer, Not a Dependency

AI is optional. Deterministic rules remain first-class. When AI helps disambiguate, Valkyrie still records the decision path so the shell does not become magic you cannot audit.