Phase 00 metric definitions — handoff note for the analysis pipeline

(Definitions only — no analysis/ code is touched by the Phase 00 chain; integration into metrics.yaml/success blocks is the pipeline owner’s call. Source: generated_reports/GNC/Jun12_26/phase00_round1_verdict.md, wiki/terminology.md.)

ze_p99 — pointing-error 99th percentile

• Signal: per-step ‖z_e − z_e_des‖ from logged tracking_error.arm.z_e and tracking_error.arm.z_e_desired (both unit camera-axis vectors).
• Units/interpretation: chord distance between unit vectors ≈ angle in radians for small errors; exactly 2·sin(θ/2) (0.86 ≈ 51°, 0.055 ≈ 3.1°).
• Reduction: 99th percentile over the operational window (post-startup, pre-completion, mode ≠ INITIAL) — same window as pe_p99.
• Direction/targets: lower is better; Phase 00 stretch < 0.001 (~0.06°). Informational (gate=None) in round 1.
• Caveat: legacy logs (pre-Jun-12 harness) lack z_e → NaN, render as an em-dash.

nu_jump_p99 — EE twist step-jump 99th percentile

• Signal: per-step ‖Δν_e‖ = ‖ν_e[k] − ν_e[k−1]‖ from logged tracking_error.arm.nu_e (6-vector twist).
• Units/interpretation: MIXED (m/s stacked with rad/s) — compare across runs only, never against a physical constant. The smoothness witness from TASK_14.
• Reduction: 99th percentile, operational window shifted by one step (op[1:]).
• Direction/targets: lower = smoother commands; round-2 rule: no >10% regression while improving pe/ze.

Reference implementation for both: validation/run_mission.py analyze() (the two metric blocks after the pe/nu computation).