Add v2_a11 commander-kill-likelihood axis from override history

Add filter axis v2_a11 (commander_kill_likelihood, 0-3) joined to v2_a1..v2_a10 composite. Compute by extracting features from the 4 known Commander overrides (3 KILL, 1 DEFER): change_type, target_files top-level dir, title token bag, presence of words like 'orchestrator' / 'meta-meta' / 'cross-vendor judge'. For each new candidate, return a score where 0 = strong KILL-like match, 3 = no match against any historical override pattern. Insert into filter_score.js after the v2_a10 axis call (around line 240 in current file). Persist feature snapshot per candidate in a new audit column so later overrides retrain the feature set offline.

Across the next 20 cycles, candidates that the engine would have produced and council-passed but Commander would have killed should score ≤1 on v2_a11 at least 60% of the time. Composite score spread between Commander-KILL-shape candidates and Commander-pass-shape candidates ≥ 1.5 points.

Leave-one-out replay over the 4 historical overrides: for each, train v2_a11 on the other 3 and score the held-out one. If the held-out KILL candidate scores ≥2 (no signal) more than once out of 4, the feature extraction is too sparse to generalize and the axis should be removed pending more override data.

Rejected by filter_score. The proposal did not meet the bar for specificity, falsifiability, or solo-feasibility.