What problem does it solve? After distilling AI session transcripts into packets, teams still need a controlled way to turn those packets into candidate memory entries for systems like Codex-mem without blindly syncing unverified claims. This Skill exports structured draft memory JSON from packets, labels each entry, and gates promotion through a human review workflow. ## Core Features & Use Cases - Packet-to-Draft Export: Parses packets/<session-id>.md and produces memory-drafts/<session-id>.json with labeled entries (new, refine, confirm, conflict, ephemeral) and readiness levels (ready-candidate, needs-raw-review, needs-conflict-review, local-only). - Review Queue Management: Supports approve, reject, defer, note, plus batch operations, writing review status, notes, and audit logs back into the draft JSON. - Sync List Generation: Derives sync-lists/<session-id>.json containing only approved, ready candidates, gated by the session's Final Session Review promotion check. - Use Case: After running session-distill on a Codex session, export the packet to memory drafts, review conflicts against existing memory, approve ready candidates, and generate a sync list for manual Codex-mem synchronization. ## Quick Start Ask the assistant to export memory drafts from the latest session packet and show the review queue so you can approve or reject each candidate.