Add forbidden-target-files clause for retired product lane to META_PROPOSER_SYSTEM

Append to META_PROPOSER_SYSTEM (after the REJECT list around line 54): 'RETIRED LANE — target_files MUST NOT begin with hypothesis_engine/moves/ or hypothesis_engine/scheduler.js. Per S173 the product lane is short-circuited (hypothesis_engine/scheduler.js:43-59 PRODUCT_LANE_RETIRED=true), so changes to those files are no-ops. ACTIVE target_files: meta_engine/*, opportunity_engine/*, layer4/*, forecaster/*. SHARED-eligible: hypothesis_engine/llm.js, hypothesis_engine/db.js, hypothesis_engine/budget_guard.js.' Then in validateProposal (line 137-172), add a guard: for each entry in p.target_files, if entry starts with 'hypothesis_engine/' AND entry is not in {'hypothesis_engine/llm.js','hypothesis_engine/db.js','hypothesis_engine/budget_guard.js'}, return { ok: false, reason: 'target_files_in_retired_lane: ' + entry }. Belt-and-braces: prompt steers the proposer; validator catches anything that slips through.

On 2026-06-12 cycle 5/5 proposals targeted hypothesis_engine/moves/* paths (council rejected 19/19 — possibly because they target a frozen lane). On 2026-06-13 only 1/5 did. Re-running with the new prompt+gate over the next 3 cycles (~15 proposals), expect ≤2/15 hypothesis_engine targets, and any that slip through are gated before persistence with reason 'target_files_in_retired_lane'.

Run 3 cycles post-change. If ≥6/15 candidates still pre-validation propose hypothesis_engine/* paths, prompt steering is ineffective — only the gate is doing the work; remove the prompt edit. If validateProposal never fires (0 gate hits across 30 candidates), the prompt edit alone is sufficient — remove the gate code.

Rejected at the council verdict. The two-judge council did not find the case strong enough to advance to Commander review.