Human review handoff
Fictional sample. Replace every metric and decision with evidence from the real review context.
Select 68% in the revised cohort, choose Create note, and use the anchored popover to reply, edit,
resolve, or reopen without a mode. The range stays highlighted; hover, tap, or focus its marker for View
thread. A selection can cross adjacent targets. Review opens only the non-reflowing list, import, and one
local version-4 review.json export for the active report language.
The bundled prior-review.json remains a valid version-2 whole-block handoff. Its legacy thread stays
list-accessible after import, demonstrating backward-compatible binding without offering new block controls.
After editing this source, rebuild with --review review.json to display truthful prior bindings and complete
the next review round without a server or automatic Markdown rewriting.
Activation evidence
Activation reached 68% in the revised cohort.
The evidence is ready for a reviewer to confirm or return for revision.
Retention evidence
Activation reached 63% in the observed cohort.
This repeated statement demonstrates that review threads remain attached to distinct source blocks.
Proposed handoff
The reviewer can resolve a thread after accepting the agent's change or reopen it with another message.
The exported review is local, deterministic, and descriptive; it is not an authenticated signature.