Under the hoodTier: how much it should matter to you
2Useful: my rating, 1 to 5
3Signal: worth watching, 1 to 5
Remote ControlArea: what it touches
Internal ChangesKind: in v2.1.268,
Internal ChangesSection of the release
Unclear The finding shows the metadata field being written but not what consumes fork_seed_sha256 or remoteControlForkChild or why.
What
Claude Code's remote-control fork handling now uses a seed value, fork_seed_sha256, to connect a forked session back to rows from its parent.
In headless mode (-p), when given an sdkUrl and credentials, Claude Code writes fork_seed_sha256 into the session's internal metadata and threads a remoteControlForkChild value through the streaming loop.
In the --print--resume path, Claude Code now asks remoteControlForkChild.adoptSeed(...) to look up rows matching that fork_seed_sha256, feeds the resulting rows in as precedingRows during session loading, and filters out any rows already adopted from what gets shown in the transcript (tracked via a liveSourceRows set). A new normalization pass runs over the loaded session data before existing checks.
Why
This lets a forked or resumed session correctly pick up conversation history from a remote-controlled parent session instead of starting without that context or duplicating rows.
Read from
Names in the bundlefork_seed_sha256
How sure we are
Something disagreesSomething we can check disagrees with this entry, or the writer said they could not settle it.
The writer flagged doubtThe finding shows the metadata field being written but not what consumes fork_seed_sha256 or remoteControlForkChild or why.
