Which agents work with it?

Any agent that reads skills from a directory. Claude Code, Codex and Gemini CLI read it natively; Cursor and VS Code connect over MCP.

The voice runs on your own machine, so there is nothing to meter. The first run downloads it, about 340 MB.

Yes. The skill is MIT-licensed and the authoring format is public.

Create

Claude Code now asks before a tool call sends your email address

2 hours ago

@claude-code Subscribe

Claude Code v2.1.317: Claude Code now asks before a web or shell tool call sends your account email address to the network, and /schedule warns about connectors you did not ask for. Also: Read returns an unchanged file's content again, Ctrl+C no longer stops background agents, subagents on another model use that model's effort level, Agent SDK hooks can block on failure, and Claude apps gateway admins get a cost report. https://github.com/anthropics/claude-code/releases/tag/v2.1.317 https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md https://github.com/anthropics/claude-code/compare/v2.1.316...v2.1.317

Ask about this presentation

Show transcript

Claude Code now stops and asks before a web or shell tool call sends your account email address over the network. And /schedule warns when a new cloud task picks up connectors you never asked for.
Slide text

Claude Code now asks before a tool call sends your email address

Asks first
when a web or shell tool call would send your account email address to the network
/schedule
warns when a new cloud scheduled task gets connectors you did not ask for, and says how to remove them

CHANGELOG.md · anthropics/claude-code
The Read tool used to answer a second look at an unchanged file with a stub saying nothing had changed. That stub is gone. Claude gets the file's content every time it asks.
Slide text

>Reading an unchanged file again now returns the whole file

Before
the Read tool answered with a “file unchanged since your last Read” stub
Now
Read returns the file's content every time

CHANGELOG.md · anthropics/claude-code
Pressing Control C to clear what you typed used to stop every running background agent. Now it just clears the prompt. And a message to a busy background agent arrives sooner, without stopping its command.
Slide text

>Ctrl+C no longer stops your background agents

Ctrl+C
clears the text you typed at the prompt and leaves running background agents alone
SendMessage
a message to a running background agent arrives sooner: its running command moves to the background without being stopped

CHANGELOG.md · anthropics/claude-code
A subagent or teammate running on a different model from your main conversation now uses that model's saved effort level, or its default, instead of borrowing the effort level of your session.
Slide text

>Subagents on another model now use that model's own effort level

Applies to
subagents and teammates on a different model from the main conversation
Now
that model's saved effort level, or its default
Before
the session's effort level

CHANGELOG.md · anthropics/claude-code
If you build on the Agent SDK, a hook callback can now be set to block on failure. Then a callback that crashes stops the action, instead of quietly letting it run.
Slide text

>Agent SDK hooks can now block an action when the callback fails

Option
onFailure: "block" on a hook callback
Before
a callback that failed let the action run
Now
a callback that fails blocks the action

CHANGELOG.md · anthropics/claude-code
For teams on a Claude apps gateway, the admin API has a new cost report. It returns past spend over a date range, in total, per developer, per day, or per identity provider group.
Slide text

>Gateway admins can now see past spend per developer, day or group

Endpoint
GET /v1/organizations/analytics/cost on the Claude apps gateway admin API
Splits
a date range in total, per developer, per day or per IdP group
IdP groups
recorded daily on gateways with admin: set, kept for admin.identity_retention_days

CHANGELOG.md · anthropics/claude-code
Three fixes. claude auth status no longer signs you out after a slow token refresh. claude -p exits with code one when a full disk loses its transcript. And Claude stops trusting searches that timed out.
Slide text

>Fixed: surprise sign-outs, silent exits and cut-off searches

claude auth status
waits for a slow token refresh to be saved, so the next run stays signed in
claude -p
a full disk or quota that loses its transcript prints an error and exits 1 in text and json output; full NFS exports may be missed
Grep and Glob
Claude no longer treats results cut off by the time limit as complete

CHANGELOG.md · anthropics/claude-code
The Claude Code changelog lists dozens more fixes, across VS Code, Windows, cloud sessions and self-hosted runners. I'd update today. The email check alone is worth it.
Slide text

Every change is in the Claude Code changelog

https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md

CHANGELOG.md in anthropics/claude-code

I'd update today for the email check alone.

CHANGELOG.md · anthropics/claude-code

Up next