[abewheel]

Schedule reviewers previously received every decision as an empty prompt, a large first-work preview, and an old narrative prefilled as today's conclusion. Review packets now distinguish unchanged dispositions from decisions requiring a fresh review and show before/after rule configuration, population, sources, allowed windows, and overdue cutoffs as of the stated date. Withdrawn excluded proposals remain visible.

The CLI prints only decisions requiring review and omits the detailed operating preview unless --include-operating-preview is requested. Generated commands create private exclusive drafts outside the checkout with --output and use compact --summary --json previews and saves; existing drafts, symlink targets and authoritative source files cannot be overwritten. Fresh reviewer, date and conclusion fields start blank. Planned prompts, including partially completed drafts, do not appear as prior completed reviews. Reviewers may explicitly reaffirm unchanged dispositions against both the exact current collection revision and prior review-record revision; the server reconstructs and validates the complete decision set from trusted historical source facts. Client packet metadata cannot authorize reuse. Existing full-array payloads and domain API preview defaults remain supported.

The browser uses the same packet, requires explicit unchanged confirmation, and preserves direct review intent while lazy state loads. Prior decisions and narratives remain historical context. No schedule, policy activation, approval identity requirement, or data model changes.

Validation: seven independent combined code/security passes converged after fixing summary counts for explicit overrides and including timezone in timing context and preserving direct-link intent across both lazy-state readiness orders; the final pass found no actionable issues; Node 24 CI=true npm run validate; focused packet, source freshness, cache, rendering, and collection-review regressions; actual 57-schedule preview/save with four changed decisions and 53 explicit reaffirmations, legacy payload compatibility, untrusted client metadata, missing/replaced prior bindings, missing historical baseline, stale source edits, reviewer conflicts, retained prior history, and unchanged activation facts. Playwright exercised the cold direct-review link with both deterministically delayed readiness orders, permanent read-only behavior, empty identity/conclusion fields, changed-window details, explicit confirmation, actual successful save, and current gate with no console errors or failed requests. The dedicated local server was stopped. Fresh generated 224-resource consumer validates without errors or warnings; both package dry runs, npm audit, and private-source/secret scans passed.

Coordinated patch release: 0.21.15, no model migration.