← all meta proposals

Restructure council_verdict.js cheapest_instant_kill_test from string to machine-readable object

filter rejected PROMPT reversible: simple 4h proposed 19 May 2026
What is the proposed change?
In council_verdict.js, change the `cheapest_instant_kill_test` field from a free-text string to a structured object in the VERDICT_SCHEMA_INSTRUCTION: `{"gate_type": "OUTBOUND_SIGNAL_CHECK" | "ARTIFACT_EXISTENCE_TEST" | "MARKET_OBSERVATION" | "BUYER_DIRECT_ACCESS" | "NOT_APPLICABLE", "specific_action": "<30 words: exact action naming platform, ICP title, and quantity threshold>", "pass_criterion": "<20 words: observable result that means proceed>", "time_bound_days": 7 | 14 | 30}`. Prompt instruction: 'Do not write generic tests. Name the platform (LinkedIn, specific Slack workspace, named subreddit), the ICP job title, and the minimum signal count. "Reach out to customers" is insufficient. "Cold DM 12 LinkedIn Heads of Support UK, need 2 replies naming missed-SLA pain within 7 days" is sufficient. gate_type=NOT_APPLICABLE only if this is a STRONG_BUILD verdict and no kill test is needed.' Apply this structure to all verdict_action values. Update the Commander dashboard to render cheapest_instant_kill_test as a structured 'Kill Gate' card with a checkbox to mark as executed.
Target files
hypothesis_engine/moves/council_verdict.js digest_server/routes/commander.js
Expected effect
Every GATHER_MORE_SIGNAL verdict (currently ~55% of verdicts per S157 distribution: 25 ROBUST + 13 MIXED + 4 FRAGILE + 1 STRUCTURALLY FRAGILE out of 43) produces a machine-readable gate condition that Commander can execute without re-reading the full reasoning. The 2 non-convergent verdicts (7199a9, 2ca131) which currently produce 'human decision required' should produce structured OUTBOUND_SIGNAL_CHECK tests that resolve the convergence dispute. Over 30 days, the dashboard will surface a priority queue of pending kill tests sorted by time_bound_days.
Falsifier — what would prove this wrong?
Apply updated council to 10 hypotheses. For every non-STRONG_BUILD verdict: gate_type must be a valid enum value (not NOT_APPLICABLE unless STRONG_BUILD); specific_action must name at least one platform, community, or ICP job title; pass_criterion must be parseable as a binary outcome; time_bound_days must be populated. If ≥3 of 10 produce specific_action fields containing only 'validate with potential buyers' or equivalent generic text, the council prompt requires few-shot examples of structured kill tests. Additionally, re-run the 2 known non-convergent hypotheses (7199a9, 2ca131): both should produce gate_type=OUTBOUND_SIGNAL_CHECK or BUYER_DIRECT_ACCESS with a testable 7-day action.
Evidence that triggered the proposal
  • Corpus E, recent council verdicts: 5d7cca 'run weekend transferability test', 26fc18 'run Week 1 outbound before building', 47730e '7-day artifact-upload test must clear', cc72cd '7-day signal check before commit' — council generates these tests as unstructured prose in every GATHER verdict; they cannot be parsed, tracked, or executed without re-reading full reasoning
  • Corpus E, recent council verdicts: 7199a9 Recruiter Claim-to-Scorecard and 2ca131 Sprint Estimate Stress-Test both 'could not converge after 3 rounds — human decision required' — structured kill_test would give Commander an executable path rather than an unresolvable escalation
  • Corpus D, S157_NBJ_DESCRIBABILITY_TEST.md: Q4 pass criterion 'Buyer can state for this specific input, correct output is X and not Y' — the cheapest_instant_kill_test is Q4 applied to validation rather than description; structuring it as an object enforces the same specificity standard

Proposer self-score

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

AxisScore
specificity3
falsifier3
solo feasible3
blast radius2
composability2
reversibility3
Disposition
Rejected by filter_score. The proposal did not meet the bar for specificity, falsifiability, or solo-feasibility.

Evaluation history

WhenMove
2026-05-19 12:54red_team_kill
2026-05-19 11:12steelman
2026-05-19 10:07meta_filter_score
2026-05-19 09:53meta_genesis