Prepare a shift handoff for interrupted evidence work — what should the operator record?
Research operations can span a shift even when the data is already returned. The next person needs the last committed boundary, not a narrative reconstructed from memory. Prepare a narrow resume note that distinguishes accepted work from an in-progress action and avoids promising that an old continuation remains usable.
Preserve the last committed operation and next safe action across an operator change.
Record the last committed table_revision and operation_key, not the last command typed or the last screen the sender remembers seeing. A committed artifact reference is stronger than a progress percentage because the next operator can compare the actual file and manifest against the ledger.
Confirm which operator owns each remaining action and the conditions that require stopping rather than improvising around an expired or blocked state. Include escalation conditions such as context mismatch or unsupported schema so the incoming shift knows when to stop rather than improvise a potentially misleading repair.Identify the safe boundary
Record the last committed table_revision and operation_key, not the last command typed or the last screen the sender remembers seeing. A committed artifact reference is stronger than a progress percentage because the next operator can compare the actual file and manifest against the ledger.
Separate pending actions
List received-but-uncommitted pages, unresolved joins and unreviewed corrections individually, with their evidence references and current states. For each pending action, specify whether evidence already arrived, parsing completed and membership was admitted; these are distinct resume boundaries with different duplicate risks.
Rehearse the resume
Ask the receiver to identify the next read-only validation step from the ledger without rerunning collection or changing permissions. The receiver should be able to say inspect pending receipt before append rather than rerun everything, showing that the note supports a concrete safe continuation.
Assign uncertain outcomes
Where an operation may have completed, require reconciliation before retrying; uncertainty should not lead to a duplicate append. If a timed-out operation might have written a local file, inspect that file and its validation state before creating another artifact under the same name.
Close the shift transfer
Confirm which operator owns each remaining action and the conditions that require stopping rather than improvising around an expired or blocked state. Include escalation conditions such as context mismatch or unsupported schema so the incoming shift knows when to stop rather than improvise a potentially misleading repair.
Keep the boundaries clear.
- Preserve the last committed operation and next safe action across an operator change.
- Operational record: Identify the safe boundary. Context badges do not certify access.
Start with a read-only check: ask your agent to report actual source coverage before collecting records or exporting an audience.
Primary references
Documentation reviewed 2026-11-15. Product capabilities, client availability and plans can change. Third-party marks identify their products, not an affiliation.