Add kill-reason concentration detector tool

council rejected TOOL reversible: simple 3h proposed 4 Jun 2026

What is the proposed change?

Standalone Node script that queries engine.db for hypotheses killed in last N days (default 7), groups by kill_reason, flags any reason exceeding threshold T (default 5 occurrences) with one-line stdout warning: '[KILL_CLUSTER] reason=X count=Y window=Nd'. Exits 0 always (advisory, not blocking). Add to scripts/ for ad-hoc invocation; document for nightly cron in README. No database writes.

Target files

hypothesis_engine/tools/kill_reason_audit.js

Expected effect

Run against current engine.db (which contains the v2_backfill_orphan_S148 cluster of 7) produces exactly one warning for that reason. Future kill-clusters surface within 24h of crossing threshold rather than at retro-review.

Falsifier — what would prove this wrong?

Run against current engine.db. If it fails to flag the v2_backfill_orphan_S148 cluster (known 7-count), OR produces >3 cluster warnings on the same baseline (false-positive rate too high), the detector thresholds are miscalibrated.

Evidence that triggered the proposal

E — engine.db trace: 7× v2_backfill_orphan_S148 kills clustered, unnoticed until retro

Proposer self-score

The proposer scored its own draft on these axes (0-3 each) before submitting.

Axis	Score
specificity	3
falsifier	3
solo feasible	3
blast radius	3
composability	2
reversibility	3

Disposition

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

Evaluation history

When	Move
2026-06-12 05:12	meta_council_verdict
2026-06-12 04:51	meta_argument
2026-06-12 04:13	meta_filter_score
2026-06-04 04:03	meta_genesis