Fetching from the wire…
01
02
03
04
05
06
07
08
Public story · 2026-07-23 · high
Fork skills chained to their own output now run before that output exists, unless they opt out.
Why now: This shipped July 22 as Claude Code's follow-up to 2.1.215, which had already stopped /verify and /code-review from firing on their own.
Claude Code made background execution the default for /code-review and any skill declaring context: fork, per the 2.1.218 release notes posted July 22. That default flip matters most for anyone who built a multi-step skill assuming a fork step blocks until it returns a result. It doesn't anymore. The next step in the chain now fires immediately, before the forked skill's output exists. A skill opts back into the old behavior by setting background: false.
This pairs with 2.1.215, which stopped /verify and /code-review from firing automatically. That release put invocation back in your hands. Running them now costs no main-thread context, since output moves to a background subagent instead of the main conversation.
The failure mode is quiet. A chained skill that reads its own fork step's output will run against nothing, or against a stale result. The error that follows will look unrelated to the frontmatter flag that caused it.
Anyone maintaining custom skills should grep their files for context: fork and check what fires right after. The declaration's meaning changed on July 22, without a rename.
Each link below shares sources, entities, or timing with this story.
If you've written Claude Code skills with context: fork in the frontmatter, go check them right now. As of the July 22 release, /code-review runs as a background subagent, and any skill declaring context: fork runs in the background by default (release notes). The opt-out is b...
Two of the six hid commands from the approval dialog itself, using tab padding, invisible Unicode, and split Bash strings.
It logs edits and errors to SQLite and retrieves them by search, skipping re-reads after compaction, enforced across 17 agent platforms.
Boundary-Bench ran 12 agent harnesses through real firewall and filesystem locks, and costs climbed as much as 167 percent as those restrictions tightened.
The open-source tool now sits at 98.9k GitHub stars and pairs a skill with a proxy that compresses code, logs and JSON before they hit the model.
The release also adds a command that migrates Python projects off the old anthropic 0.x SDK and stops WebFetch from serving stale content all session.
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.