[merynz]

Mainline consolidation — first bounded integration slice (Issue #361)

Import only reviewed, unchanged V6 typed contracts, from PR #297 exact head dd3089957e3126d93db4ad2804e30511cfdca348:
rig_program_ir_v1.py, rig_constraint_schema_v1.py, skin_code_ir_v1.py,
canonical_rig_ir_v2.py and canonical_skin_ir_v2.py.

Their upstream main dependencies (canonical_mesh_ir_v1, carrier_skin_v1, hashing, types) have identical blob SHAs on V6 and current main.

Includes the original V6 positive/negative compiler tests plus new main-specific
Skin V2 hash tampering and unsupported operator tests, an independent Actions
lane, and scope/authority document.

Explicitly no product worker, Go registration, V7 neural models, Mechanics
solver promotion, Corpus370/training, or Graphics changes. This
imports contract code, not a ready-to-ship compile or ModelRelease.

Review and CI first; do not auto-merge. Following success, make short
main-based runtime/worker/behavioral PRs rather than bulk-merging
the stale 137-commit #297 branch. Track under #361.

[merynz]

Review verification — exact preserved source + main-bound behavior

Checked every copied V6 file against the original, pinned V6 commit dd3089957e3126d93db4ad2804e30511cfdca348: 7/7 Git blob SHAs identical (5 schema/IR sources + 2 unchanged original V6 regression files). The 4 shared imported runtime dependencies have identical V6/main blobs.

PR head 859c899258dd07a25480208adf1fe8263d38e7a8. Dedicated CI 42297491225 completed 13 passed. The two extra main-focused tests exercise a full synthetic RigProgram→CanonicalRig.v2→qualified carrier skin→CanonicalSkin.v2 mint, and independently reject a correctly hashed but wrong deform-structure binding.

Research/source firewall: no new inference, no V7 trained ModelRelease, no Mechanics execution, no ProductRevision authority, no Graphics/Corpus changes. Draft—do not merge until the full review/required checks settle.