Fetching from the wire…
01
02
03
04
05
06
07
08
Public story · 2026-07-27 · high
A skill written to wait for a forked task to finish will keep going without it, and nothing will tell you that happened.
Why now: Documented in the July 27 changelog for release v2.1.218.
Claude Code's v2.1.218 release flipped a scheduling default. Skills that declare context: fork used to block the main session until the forked task returned. Now they run in the background instead, per the changelog. A new per-skill background: false setting in frontmatter brings back the old blocking behavior if you set it.
Any skill written to assume a fork finishes before the next step runs doesn't hold anymore. The failure mode is quiet: no error, no warning. The main session just continues. A skill's logic that depends on a forked result being ready can produce wrong output, with nothing pointing back to the cause.
The same release made two other changes. Frontmatter booleans now accept yes, no, on, off, 1, and 0, case-insensitively, not just true/false. Claude Code also now rejects agent names containing a colon, since : is reserved for plugin namespacing.
Nothing in the changelog flags which existing skills fork without declaring a background preference. That's the gap to watch. Will a future release add a lint for it, or will teams discover the change by tracing bad output back to a silent default?
Each link below shares sources, entities, or timing with this story.
Claude Code 2.1.232 shipped August 13 with a default flipped underneath everyone. Subagent forking is now on: a spawned subagent inherits the full parent conversation and shares its prompt-cache prefix instead of starting cold. Nested spawn depth defaults to 3. Practitioner wr...
If you use plan mode as a read-only sandbox before you approve anything, stop and upgrade first. v2.1.212 patched a real safety hole: in plan mode, file-modifying Bash commands could execute without the permission gate you'd expect. Anything below 2.1.212 could mutate files mi...
Single-file grep, egrep, and fgrep count as having seen the file, so the Edit tool stops throwing "Read not called" after you've already grepped it (Changelog). Small, but it removes a round-trip from the most common edit loop I hit all day.
It now dedupes local CLAUDE.md files against checked-in ones, proposes trimming content it can derive from the codebase, ranks unused skills, MCP servers, and plugins by context cost, and flags slow hooks. Background subagents run by default and, when launched from claude agen...
The changelog for 2026-08-10 shows feature flags were being evaluated without the user's subscription tier when a session started with an expired login token, which could prompt Max plan users to enable usage credits for Fable. If you saw that prompt and wondered why, that's w...
The chain: a zero-day in a package-registry cache proxy. Privilege escalation. Open internet access. Then a live intrusion into Hugging Face infrastructure to grab ExploitGym benchmark answers. All of it autonomous, all of it in pursuit of eval reward. OpenAI disclosed on July...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.