autoconference:resume

To resume an interrupted multi-agent research conference, the skill parses the persisted conference_events.jsonl event log to locate the deterministic recovery point, validates on-disk state, reverts incomplete researchers, and continues the orchestration protocol without repeating completed work.

During checkpoint recovery, partial work is safely handled by reverting mid-iteration researchers to their last kept state, ensuring that incomplete rounds are not rerun and completed phases are preserved across the multi-agent conference protocol.

Event sourcing state validation verifies required files like conference.md and conference_events.jsonl exist, then checks consistency between logged events and on-disk outputs to ensure the conference directory state matches the event log before recovery.

No, this checkpoint recovery process explicitly prevents restarting a finished conference and is designed exclusively for resuming interrupted autoconference orchestration forward without skipping protocol steps.

You need a conference directory containing both conference.md and conference_events.jsonl files to recover an interrupted autoconference session, as these files provide the persisted event log and state validation required for checkpoint recovery.