CREATURE_BODY_PLAN_TAXONOMY.md

creatures/CREATURE_BODY_PLAN_TAXONOMY.md

The creature body-plan taxonomy — how many skeletons exist

Proposal tier. Authored 2026-08-13. Named for Josh's veto per the standing rule that

beyond-floor additions are declared rather than slipped in.

Why this is the first thing, not the forty-second

T0_Creature_Roster classifies creatures biologically: 82 rows say mammal. A bat, a whale, an

elephant and a rat are all mammals and share no rig, no skeleton, no animation, and no scale. 47

rows say folklore, which describes provenance and not a body at all. There is no body-plan column

on the registry, and height_m is populated on 7 of 187 rows.

So the roster today cannot drive a rig or a scale. That is the actual blocker under the creature

lane, and it gets worse with volume: the canon requirement is roughly 2,000 species

(docs/ASSET_BUILD_LIST.md). Assign body plans well and that is ~30 skeletons with ~2,000 meshes

hung off them. Assign them badly, or not at all, and it is 2,000 rigging problems.

A body plan is not biology. It is limb topology plus spine articulation plus ground contact

the three things a skeleton encodes. Two animals in different orders share a plan when they share

those; two animals in the same order do not when they don't.

The taxonomy — 30 plans

Terrestrial vertebrate

leopard, clouded leopard, sun-bear-adjacent felids and canids.

pig, rhinoceros.

plantigrade because the gait solver and foot IK differ fundamentally.

small mustelids.

langurs.

This is the highest-leverage row in the table: a hominin cryptid rides the same rig as a

character, so the entire human animation library and the MetaHuman retarget path reach it for

free. Scale differs; topology does not.

Flying

Its own plan because neither parent plan animates it.

Nothing in the feathered plans transfers.

Reptile

Aquatic

living fossil whose whole point is that its fins are almost limbs, and the standard fish rig

destroys exactly that.

Invertebrate

Folklore composites that need their own rig

These are not a "folklore" bucket. Each earns a plan because its silhouette cannot be animated by

any plan above.

the seven-headed balustrade water-lord.

Loch Ness form, Bunyip's aquatic read.

cadence all change. Thunderbird, Ropen.

Genuinely unlike anything else in the table and it needs a bespoke simulated chain.

static or semi-static, animating from a stone rest pose rather than a creature gait.

What this buys

withoutwith
skeletons to author across ~2,000 speciesup to 2,000~30
animation setsper creatureper plan
creature rows reachable by the human/MetaHuman rig0every hominin_biped

The two rules that keep it honest

1. A plan is claimed by topology, never by culture or provenance. "Folklore" is not a body

plan. A folklore being gets the plan its described body actually has, and if it has none, it

earns a new plan and the new plan is named here.

2. A row with no cited description does not get a guessed plan. It goes to UNASSIGNED with

the missing canon named. A wrong plan is more expensive than an empty one, because a wrong plan

is discovered after the rig is built.

Care note

Assigning hominin_biped to Ebu Gogo, Orang Pendek and Nittaewo is a statement about **skeleton

topology for rigging purposes only**. These are beings in living oral traditions and, in the Ebu

Gogo case, adjacent to a real palaeoanthropological record on Flores. The plan column says how the

mesh articulates. It says nothing about what they are, and it never routes their depiction — that

stays with the region page, the design sheet and the care tier.

Generated by harness/site/structure_site.py — the URL path is the repo path. review root