agentleFS
jammyfujammyfu/open-game-skills/skills/disciplines/animation-blend/SKILL.md

animation-blend

Skill13 starsChanged 24 days agoMITView on GitHub

Use when rendered character poses need weighted locomotion, aim, attack, additive or transition blending while gameplay state, hit timing and displacement ownership remain explicit.

The file

We split this skill into the rules it gives an agent

1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Before tuning, attribute jitter by measuring rendered joints with the corrective layer on and off.

    line 34

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Animation blend

This skill owns presentation pose composition, not gameplay state or hit timing. animation-graph selects/coordinates animation states/layers; movement owners define displacement policy.

Compatible displacement columns

ColumnGameplay displacement owner
procedural-locolocomotion/physics code
root-motionauthored root motion consumed by the gameplay movement solver
hybridpublished state-dependent ownership with one displacement application

Rules

  1. Weighted blends/additives/interpolation paint a rendered pose from authoritative animation/gameplay state; they do not extend active hit frames or cancel windows.
  2. Blend duration/easing/inertialisation are authored in time/state units appropriate to the runtime, not a universal 2–6 frame fade.
  3. Root-motion extraction/application has one explicit gameplay owner. Visual blending must not double-apply root displacement.
  4. Aim/look overlays declare mask/weight/reference space and do not silently rotate gameplay facing/collision unless the owning gameplay contract requests it.
  5. Foot contact repair delegates to ik-foot-locking; blend output remains a source pose rather than a second locomotion solver.

Corrective layers

Procedural corrections applied after the source pose are presentation layers with the same ownership limits. Examples are limb/body interpenetration avoidance, reach or look IK, and prop alignment. They must also be temporally stable:

  1. The correction is a continuous function of pose and time. Avoid hard thresholds, snapping to bins or cells, picking whichever sample is deepest, and hard on/off zones. Blend across regions and weight competing samples smoothly.
  2. Correct through a degree of freedom that can actually remove the contact. For a contact on the upper segment, rotate the chain about its parent joint rather than pushing an end effector that barely moves that segment. Iterate with bounded gain and no overshoot; a case that does not converge must not alternate between frames.
  3. Contact onset is soft enough that brushing contact causes no velocity kink. The allowed residual overlap is an authored tolerance, validated against the project's clipping check.
  4. For deterministic clips, corrections may be cached and filtered over time per actor and clip. The cache is bounded, is filled away from frames that already carry heavy work, is invalidated on rig, body or clip changes, and is released with the actor.
  5. Before tuning, attribute jitter by measuring rendered joints with the corrective layer on and off. Fix step thresholds in the source motion separately.

Accept

Record animation/gameplay state, source clips/layers/weights, displacement owner and resulting pose through transition/interruption cases. Disabling render meshes/blending does not change authoritative hits, cancels or gameplay displacement.

For corrective layers, sample joint positions at the presentation rate. Count consecutive-frame acceleration reversals, or another authored jitter metric, with corrections on and off, alongside the penetration check. Neither metric may regress.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.