{"schema":"open-scaffold.evolution-attempt.v1","attempt_id":"attempt-a","recorded_at":"2026-05-22T10:10:00.000Z","run_id":"attempt-a","task_id":"task-reviewable-csv-importer","run_packet":".osc/runs/attempt-a/run.json","evaluation_id":"eval-attempt-a","evaluation":"docs/evidence/attempt-a-evaluation.json","evaluation_decision":"approved","decision":"promote","score":0.62,"rationale":"First useful baseline; AC2 still returns a generic error. Promoted to set the frontier so improvement attempts can be measured against it.","evidence_refs":["docs/evidence/attempt-a-proof.md","docs/evidence/attempt-a-evaluation.json"],"adapter_receipts":[],"boundary":{"runtime_spawning":false,"model_benchmarking":false,"compliance_certification":false,"approval_or_release_decision":false,"external_anchoring":false}}
{"schema":"open-scaffold.evolution-attempt.v1","attempt_id":"attempt-b-prompt-rewrite","recorded_at":"2026-05-22T10:20:00.000Z","run_id":"attempt-b-prompt-rewrite","task_id":"task-reviewable-csv-importer","run_packet":".osc/runs/attempt-b-prompt-rewrite/run.json","evaluation_id":"eval-attempt-b-prompt-rewrite","evaluation":"docs/evidence/attempt-b-prompt-rewrite-evaluation.json","evaluation_decision":"approved","decision":"reject","score":0.54,"rationale":"Rejected. Prompt-rewrite approach regressed AC3 and did not fix AC2. Filed in the ledger so the next attempt does not retry the same approach.","evidence_refs":["docs/evidence/attempt-b-prompt-rewrite-proof.md","docs/evidence/attempt-b-prompt-rewrite-evaluation.json"],"adapter_receipts":[],"boundary":{"runtime_spawning":false,"model_benchmarking":false,"compliance_certification":false,"approval_or_release_decision":false,"external_anchoring":false}}
{"schema":"open-scaffold.evolution-attempt.v1","attempt_id":"attempt-c","recorded_at":"2026-05-22T10:30:00.000Z","run_id":"attempt-c","task_id":"task-reviewable-csv-importer","run_packet":".osc/runs/attempt-c/run.json","evaluation_id":"eval-attempt-c","evaluation":"docs/evidence/attempt-c-evaluation.json","evaluation_decision":"approved","decision":"promote","score":0.94,"rationale":"Promoted to current frontier. AC2 now reports row and column of the offending token; AC1 and AC3 remain passing. Operator promoted after manual review of the evaluation envelope.","evidence_refs":["docs/evidence/attempt-c-proof.md","docs/evidence/attempt-c-evaluation.json"],"adapter_receipts":[],"boundary":{"runtime_spawning":false,"model_benchmarking":false,"compliance_certification":false,"approval_or_release_decision":false,"external_anchoring":false}}