root-motion skill

Overview

This skill provides root motion support to keep agent movement and animation synchronized with physics and navigation. It ensures character displacement driven by animation is applied reliably to the agent's transform so movement, collisions, and path-following remain deterministic. The skill is designed to integrate into Babysitter-managed workflows for reproducible, resumable orchestration of movement logic.

How this skill works

The skill inspects animation data and extracts root delta transforms each frame or tick, then applies those deltas to the agent’s position and rotation while respecting navigation constraints and collision responses. It exposes hooks to pause, blend, or override root-driven movement, and offers deterministic application order so repeated runs produce identical results under the same inputs.

When to use it

• When animations include translation/rotation baked into the root bone and you want the world transform to follow accurately.
• For synchronized movement across physics, animation, and navigation systems in agentic workflows.
• When you need deterministic, resumable movement for testing, replay, or distributed orchestration.
• To avoid slipping, clipping, or drift between visual animation and collision/grounding logic.

Best practices

• Extract and apply root motion deltas in a single, deterministic update step to avoid order-dependent differences.
• Provide clear blend modes (full, partial, or additive) and expose weights for smooth transitions.
• Clamp or validate deltas when integrating with navigation meshes to prevent invalid positions.
• Record root motion inputs and outputs for resumability and reproducible debugging.

Example use cases

• A game agent driven by complex locomotion animations that must remain on a navmesh during deterministic simulation runs.
• Automated test harnesses that replay movement scenarios exactly for regression testing.
• Distributed orchestration where multiple agents’ movement must be paused, resumed, and reconciled across nodes.
• Animation-driven interactions (pushes, grabs) where collision response must align precisely with visual motion.

FAQ