Verify a decision

{"actor":"judge","decided_at":"2026-08-27T10:40:22.787Z","decision":"review_revise","investigation_id":"2a48b3c5-5a6f-4913-a093-0800cc23a7e9","new_score":33,"page_slug":"bonk","prev_score":45,"reason":"The numeric disputed rate (1 disputed + 0 unverifiable out of 11 claims = 9.1%) falls in the approve band on its own, but two of this run's explicit adjustment rules override a flat approval here. First, claim_findings[7] disputes the page's own summary statement that there are 'no protocol-level exploits recorded' -- the reviewer found a well-documented ~$20M BonkDAO governance-attack treasury drain (~July 2026) and a subsequent Upbit delisting citing unresolved security issues, neither mentioned on the page. This is not a peripheral fact; it directly contradicts the core safety assurance a risk-intelligence page exists to provide, on a widely-held token, so it carries more weight than an ordinary single disputed item. Second, coverage_gaps lists a high-priority gap for exactly this same missing security-incident history, and this run's rules direct that a high-priority gap should push a page toward review_revise even when the disputed percentage is low. Compounding this, claim_findings[5] shows the page presents a 1-trillion-token burn as a hypothetical future condition when the threshold was actually reached and the burn scheduled/executed roughly eight months before this review (marked stale, wrong-direction). Given these findings, together with the reviewer's moderate (0.75) confidence, a light downward revision is warranted rather than either a clean approval or a denial.","score_delta":-12,"sequence_num":3,"submission_content_hash":null,"submission_id":null,"submission_kind":null,"submission_valence":null,"v":1}