Unclear What the earlier check returned is not known, so whether this changes anything for users who already had it on is unclear.
What
The Skill tool (which runs a skill, a packaged set of instructions Claude can follow) now always keeps its results when it runs inside a subagent, a helper copy of Claude given part of a task. Before, this depended on a check made at run time; it is now fixed on.
Two smaller changes come with it:
- When a skill cannot be found, the diagnostic report now includes the command name for Claude in Chrome.
- Skills whose names contain
:can fall back to a built-in skill passed in when no other match is found.
Why
What the Skill tool returns inside a subagent no longer varies, so subagents always see the result of a skill they ran.
Something disagreesSomething we can check disagrees with this entry, or the writer said they could not settle it.
The writer flagged doubt
What the earlier check returned is not known, so whether this changes anything for users who already had it on is unclear.