agents · review · first seen Nov 11, 10:32 UTC

Block releases goose v1.59.0 with voice and session updates

Goose v1.59.0 shipped per-start voice instructions for live-voice and client-tag filtering for ACP session/list. Builders using either feature have a reason to check whether the new controls fit their workflow.

L0 · primary1 source2 claims checked4 min read

What shipped

The v1.59.0 release notes list configurable, per-start voice instructions for live-voice. The per-start part matters: the release identifies a way to set instructions for an individual start, rather than describing only a general voice configuration. The excerpt does not show the setting’s name, its format, or an example. It establishes that the feature shipped, but not how a particular instruction will affect a live-voice interaction. [1]

The same release lists filtering of session/list by client tags in ACP. That is a specific change to session listing, not a claim that tags alter sessions or govern access to them. For a builder who already uses client tags, the question to test is whether the returned list can be narrowed to the sessions they intend to inspect. The excerpt does not specify filter syntax, matching behavior, or what happens when a session has no relevant tag. [1]

These are separate controls in one shipped release. One concerns instructions supplied when live-voice starts; the other concerns which sessions appear in an ACP session/list response. The release excerpt does not say that either feature requires the other, or that configuring voice changes session tags. Treat them as two checks against existing workflows, not as a combined voice-and-session feature. [1][2]

Why it matters if you build

For a live-voice workflow, per-start instructions give a builder a configuration point at the moment a voice interaction begins. That may be useful when the intended instructions differ between starts. Before adopting it, check whether the available configuration can express the instructions your workflow needs and whether the behavior is observable in your own test. The release excerpt confirms the control exists; it does not establish instruction precedence, persistence, or reliability across starts. [1]

For an ACP workflow that lists sessions, client-tag filtering offers a way to request a narrower list. That can make inspection more targeted when the client tags in use distinguish the sessions a builder is looking for. It is still worth comparing a filtered list with the list your workflow previously received. The release excerpt does not establish whether the filter changes pagination, ordering, or other session/list behavior, so those details should not be assumed. [1]

Neither release-note entry gives a measured cost, latency, or resource impact. There is no basis here to promise a cheaper voice interaction or a faster session query. The practical decision is narrower: determine whether per-start voice instructions or client-tag filtering solves a current configuration or listing problem, then verify the behavior in the relevant workflow before depending on it. [1][2]

How to try it today

1. Review the v1.59.0 release entry for the live-voice and ACP session/list changes.
2. If you use live-voice, locate its per-start voice-instruction configuration in v1.59.0.
3. Start a live-voice interaction with an instruction suited to your test, then check the observed behavior.
4. If you use ACP session/list and client tags, request a list filtered by a tag you use.
5. Compare that response with the sessions you expected to find; repeat with another relevant tag if needed.

What we checked

Every claim below was walked down the source ladder. The rung is how close the source is to the thing itself; the tense is the source's own.

1. shipped L0 · primary

   Goose v1.59.0 adds configurable, per-start voice instructions for live-voice.

   per GitHub releases · block/goose (L0)
2. shipped L0 · primary

   Goose v1.59.0 adds filtering of session/list by client tags in ACP.

   per GitHub releases · block/goose (L0)

What is not confirmed

The supplied release excerpt does not give the voice-instruction option name, accepted format, precedence, or examples. It also does not document the client-tag filter syntax, matching rules, behavior for untagged sessions, or compatibility requirements. It provides no measurements of cost, latency, or reliability for either change. Those points need verification before they become operational assumptions. [1][2]

Who should act, who should wait

Act now if you already use live-voice and need instructions that differ by start, or if you use ACP client tags and need to narrow session/list results. Make the first action a small test, because the release excerpt names the controls without specifying their detailed behavior. Wait on a workflow change if you need documented precedence, exact filter semantics, or measured performance before adopting either control. [1][2]

Sources

• v1.59.0 — GitHub releases · block/goose L0 · primary