← all meta proposals

Add pre-filter shape gate: drop audit/validator proposals lacking engine-output falsifier

council rejected GATE reversible: simple 4h proposed 13 Jun 2026
What is the proposed change?
In runFilterScoreOnOne (around line 130), before scoreProposal(), check the parsed metaDesc.title + metaDesc.proposed_diff (lowercased) against SHAPE_KEYWORDS = ['audit', 'validator', 'health check', 'observability', 'lint', 'monitor', 'metric collector', 'dashboard']. If any keyword matches, ALSO check whether metaDesc.falsifier text contains at least one OUTPUT_TOKEN from a hardcoded list: ['composite_rank_score', 'meta_ship_status', 'verdict_action', 'verdict_score', 'filter_score_median_total', 'killed', 'meta_filter_score move', 'hypotheses table', 'moves table']. If shape-keyword present AND no output-token present in falsifier, short-circuit: record a meta_filter_score move with output {gate:'audit_shape_no_output_falsifier', verdict:'DROP'} and route through runTransition to meta_ship_status='filter_rejected'. No LLM call.
Target files
meta_engine/moves/filter_score.js
Expected effect
3 of 4 historical Commander overrides were on audit-shaped meta proposals. The gate would have caught the same shape pre-LLM, saving ~$0.013 downstream per gated row + Commander review slot. Replay against last 19 proposals: 1-3 should match (the historical rate).
Falsifier — what would prove this wrong?
Run gate against 19 historical proposed rows in dry mode. If 0 match, keywords are too narrow — the gate is dead code. If ≥5 match (false positive sweep catching legitimate work like the AXIS proposal here), keywords are too broad — narrow or remove 'metric'/'monitor'. Target band: 1-3 matches.
Evidence that triggered the proposal
  • E — Engine traces: 4 Commander overrides (3 KILL + 1 DEFER), pattern = audit/observability-shaped
  • D — brain/MANIFESTO.md v4.0 (S183) — deflated to honest TBD moat, no compliance/audit infra pre-paying-customer
  • D — S177 Foundry Sentinel doctrine — internal-first metacognition, SENTINEL product hibernated; same shape leaks into meta proposals

Proposer self-score

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

AxisScore
specificity3
falsifier3
solo feasible3
blast radius3
composability2
reversibility3
Disposition
Rejected at the council verdict. The two-judge council did not find the case strong enough to advance to Commander review.

Evaluation history

WhenMove
2026-06-13 04:21meta_council_verdict
2026-06-13 04:13meta_argument
2026-06-13 04:07meta_filter_score
2026-06-13 04:04meta_genesis