cc-anticipation: Project Charter

**Document Version**: 0.1.0 **Created**: 2025-12-26 **Status**: Phase Zero - Active **Canonical Input**: [Anchor.md](../../../../docs/Anchor.md) cc-anticipation converts stabilized motion state into actionable anticipatory signals. It answers one question: **"What futures are cheap vs expensive given what's happening now?"** It does not generate motion. It does not generate sound. It does not decide policy. It **measures the geometry of possibility** around the present moment. | Responsibility | Description | |----------------|-------------| | **MotionWindow consumption** | Accept aligned, canonical motion windows from cc-window-aligner | | **Kinematic feature computation** | Compute velocity, acceleration, jerk, coordination from skeleton | | **Latent dynamics computation** | Compute dz/dt, d²z/dt² from LIM-RPS latent stream | | **Regime embedding** | Map motion window to regime embedding (64-256 dims) | | **Scalar signal emission** | Emit commitment, uncertainty, transition_pressure, recovery_margin, phase_stiffness, novelty, stability | | **Constraint proximity** | Compute balance margin, joint limit proximity, rotation commitment | | **AnticipationPacket emission** | Emit frozen-schema packets every frame, deterministically | | **Neighbor-based uncertainty** | Use continuation dispersion from MotionPhrase library | | **Deterministic replay** | Same input → same output, always | | Excluded | Rationale | |----------|-----------| | Raw sensor ingestion | Handled by cc-mcs-headless, mocopi-udp-ingest | | Time alignment | Handled by cc-window-aligner | | Policy decisions | Handled by Conductor | | Motion generation | Handled by cc-motiongen | | Audio generation | Handled by MotionStrudel, EchelonDiT | | Semantic labeling | No "spin coming" labels; only actionable scalars |