seed-a-lanelisted

# Seed a Lane (attended, env-gated) New strategies enter the forward record ONLY as new pre-registered lanes with their own config hash and inception. This is the procedure that has now shipped mirror/conviction (P1 #6) and conservative-atr (TRIAL-EXIT) safely. ## Why this exists (paid-for lessons) - **2026-06-17:** a seed flag was flipped while the wiring didn't exist — the seed was a silent NO-OP discovered only via the registry dump. Build and test the wiring BEFORE asking for the flag. - Timing artifact same day: "registry still shows N" can be a race with the background seed's commit — re-query before declaring failure. - In-place YAML edits reuse old content hashes and corrupt segment identity. ## Procedure 1. **Pre-register first** (see pre-register-trial): hypothesis, decision rule, canonical doc in `docs/TRIALS/`, committed BEFORE any forward data accrues. 2. **Own config file, own hash:** the new lane gets its own YAML (pattern: `data/conservative_atr_lanes.yaml`) — never appended into a live config whose hash identifies existing segments. Mandates copied byte-identical where the trial demands a matched control. 3. **Build the full wiring, no-op until seeded** (mirror the exit-lane pattern, `exit_lane.py`): idempotent `seed_*` function that registers the trial on seed; env-gated hook in `main.py` (`AEGIS_SEED_<NAME>=1`); scheduler daily/MTM/nav-freshness, track-record router, and registry N_eff all skip-until-seeded. 4. **Test befo