IDs and AI-friendly rules
Every item has a unique ID with a layer prefix, and every rule is written so a machine can check it.
ID prefixes
LayerPrefixExample
Semantic principlesp.sp.4
Guardrailgr.gr.repeat_max_2
Combination rulecmb.cmb.H2
Operatorop.op.antic
Channel principlech.ch.face
Rolerole.role.attend
Tokentok.tok.body.tilt.m
Behaviorcmp.cmp.curious
Situationpat.pat.calling
Form factor · source · robotff. · src. · robot.ff.legged
Stylestyle.style.animation
Evidenceev.ev.b11.malfunction
Seven rules for AI-readable data
- Every item has a unique ID with its layer prefix.
- Body parts are role-based.
- Units are normalized (see Units and evidence).
- Rules are checkable expressions (check).
- Each rule has a level MUST / SHOULD / MAY and on_violation: clamp / substitute / skip / warn.
- Two-way: making (recipe) and recognizing (recognize).
- Evidence tags (evidence).