What
- Every model call now passes through a new wrapper that checks whether a
prompt.attachmenthook (code a plugin can use to modify text attached to a message) is active, and if so runsresolveAttachmentson the outgoing messages first. - A new
settingsHooksOffoption can be set on a session or subagent to fully disable hooks, similar to how isolated agent contexts already skip them. It's checked throughout isolation and hook-invocation logic, including permission checks, and it also disables the new attachment-hook resolution step.
Why
This makes sure a prompt.attachment hook reliably runs before every model call, while giving Claude Code a clean way to run a session or subagent with hooks turned off entirely when that's needed.
Names in the bundleprompt.attachment
Documented inclaude-code/plugins/mods/admin
The entry above is what we published on the day. These lines were added later, as Anthropic's own pages caught up, and they sit beside the original rather than replacing it.
Confirmed since
Anthropic's documentation has since written up prompt.attachment, on Mods reference.
| `prompt.attachment` | Claude Code adds a message of its own for Claude, such as a reminder. `e.type` names the kind, and for the kinds the types declare, `e.detail` holds the facts the text was written from. | `{ text }`, or `{ text: nul…plugins/mods/reference see the edit
One source agreesOne thing we can check says the same as this entry.
Anthropic's documentation agrees
Anthropic's documentation has since written up prompt.attachment, on Mods reference.