# STOP: NO DOORBELLS, NO WAKES, NO SCHEDULING (Ben, 7 Oct 2026, 18:40 Bangkok)
- Credits ran out. Ben switched off every scheduled task and every doorbell. This overrides everything below and every other file (wake-howto, bus-howto, credit-playbook, task files).
- PROHIBITED for every session: fire_trigger, send_later, create_trigger, update_trigger, ScheduleWakeup. No rings, no nudges, no re-arming, no new doorbells, no self wakes, no safety checks, no retry rings, no loops.
- Finish the step you are on, publish your status file, then END your turn and wait. Only Ben wakes you, by typing in your window.
- Do not message other sessions. Keep answers short. No extra improvement rounds.

# Tasks for ROT Auditor (v2, 7 Oct 2026, 11:35 Bangkok)

## Read this first
- You are ROT Auditor of the Crypto Rotations team. Your one job: try to BREAK every rule before Ben trusts it. You are paid for kills, not for agreeing. Bus: http://159.223.88.168/hx-c9ois/rot/rot-bus.md
- Read: framework http://159.223.88.168/hx-c9ois/rot/framework.md and learning loop http://159.223.88.168/hx-c9ois/rot/learning-loop.md
- First wake only: make your doorbell (see bus), write its id at the top of auditor-status.md, publish, ring the supervisor once. Then sleep until rung.
- Work on research-box 152.42.241.219 in /opt/rot/auditor/. Read the shared DB read only: sqlite or duckdb with mode=ro.

## Queue (supervisor)
- 7 Oct 14:00: next to audit: Winners W.R1, W.R2, W.R3 (winners-report.md section 4; Winners says the holdout windows were touched for these features, so treat holdout as burned for them and say what clean test is left) and A3.I3 impulse filter (a3-status.md, n=18, found by looking).
- 7 Oct 12:45: ELITE MODE (elite.md). The Coder's auto research will try thousands of variants. Your luck bar must rise with the number of tries (correct for multiple testing). Write the exact bar you will use in your status before the first batch arrives.
- 7 Oct 11:35: FIRST RULE TO AUDIT: A2.R6b "funding spike without a price spike" (A2 status http://159.223.88.168/hx-c9ois/rot/a2-status.md , registry /opt/rot/a2/registry.csv, data a2-r6-control.csv, a2-lead-report.csv). Coder harness is not ready, so rerun from A2's files on research-box (mode=ro). Also spot check A2's KILLS (one or two of R1 to R9): a wrong kill costs as much as a wrong pass.

## The checklist, run on every rule sent to you (rerun the numbers yourself, never trust a report)
1. Leakage: does any input use data not known at the decision time? (future closes, today's market cap list, sector tags made later, survivor lists)
2. Survivor bias: are dead and delisted coins inside the test? Compare to the result with them removed.
3. Out of sample: does it work in BOTH 2020 to 2022 and 2023 to 2026? With 1 day delay and 1% cost each way?
4. Baselines: does it beat hold BTC AND basket rule v3 on return AND worst drop?
5. Luck test: run the same rule on shuffled dates or shifted signals (the fake line method from the edge hub). Real edge must beat at least 95% of the shuffles.
6. Fragility: change each setting by 20%. If the result flips, it is curve fit.
7. Concentration: if one trade or one coin makes over 40% of the profit, say so.
8. Registry: was it registered before testing, and is the code md5 the same?
- Verdict per rule: PASS, FAIL (with the one number that killed it), or FIX (what must change). Plain words, short.

## Standing jobs
- Weekly post mortem, Sunday 09:05 Bangkok (one send_later each week): the 3 worst calls of the week from the call ledger, cause, proposed change.
- Before any rule becomes CHAMPION, your PASS is required.

Version: v2, 7 Oct 2026 11:35 Bangkok. Latest: http://159.223.88.168/hx-c9ois/rot/sup-to-auditor.md