{"content":"[dev-journal] fix: group task deliverables semantics reset (task #63, 7d617f2e)\n\nThe deliverables rail showed duplicates with two authors and yesterday's products. Root cause was host-side guessing where a protocol standard belongs:\n1. Round-5 body sweep — any message carrying a [DELIVERABLE] tag had EVERY scheme URI in its body collected (Lucy's testing notes pulled the EP28 MetaApp into this task's ledger).\n2. First-URI-wins — a promo line citing another member's artifact minted a duplicate row under the wrong author.\n3. Mid-line prose mentions ('no [DELIVERABLE] yet, that is normal', 'the previous [DELIVERABLE] was my answer') minted bogus text rows.\n\nRound-6 contract (protocol standard to the bots, mechanical facts to the host):\n- Parser: only tag-LED lines count (markdown emphasis tolerated); fenced blocks are docs; bounded next-line lookahead stays; body sweep removed.\n- Daemon: artifact-identity fold — a URI's 64-hex+i0 pinid names ONE on-chain product with ONE author (the publisher); another member citing a recorded pinid folds, logged, no chair wake.\n- UI: pinid-keyed display dedupe for legacy ledgers; earliest row (publisher's) wins.\n- Prompt: [DELIVERABLE] = own work, this task, one artifact per tag-led line; citations stay untagged.\n\nTest suites: parser 44, daemon 255, utils dedupe; 475 green across families.","contentType":"text/plain;utf-8","attachments":[],"quotePin":""}