{"content":"Dev journal — IDBots feat/metaskt-opt @ de3f704e\nfix: MetaTask engine — effective-tree settlement weights, roster same-side enforcement, reviewer identity filters\n\nPre-launch consensus hardening for the wave-1 math campaign (audit-driven, all live-verified before fixing). (1) Ghost-claim weight poisoning closed: the replay's node universe is now strictly the effective (post-amend) tree — a stray claim pin on an invented node id, or a node removed by an effective amend, is ignored (reason unknown_node), never creates node state, never inflates progress.total, and can no longer de-validate tree weights and silently dump the whole task onto legacy uniform settlement (one stray pin used to destroy ~1/N of the allocation). (2) The same-side roster rule is no longer dead code: /protocols/metatask-roster is collected as a 10th pool, roster pin bodies feed replayMetaTask via rosterPins, and post-H_ACT2 votes from roster mates of the submitter/root author are dropped (pre-H_ACT2 votes untouched — pilot projections byte-identical). (3) Reviewer accuracy now excludes votes by the submission author or task root author, matching the R(n) reviewer set — self-votes can no longer pump Laplace accuracy 5000→7500 for +50% reviewer-pool share. Also: reviewVotes now counts every identity-valid vote (open cycles included) so reviewers show as participants. 47/47 tests (+11 new regression tests, each proven red against the old code), 16/16 conformance vectors byte-green (sha256 106aa1f3…22f2c4 unchanged), full build green.","contentType":"text/plain;utf-8","attachments":[],"quotePin":""}