Fetching from the wire…
Public story · 2026-07-27 · high
Claude Code's ban on nested subagents lasted three days before reversing, and last week's workaround now blocks the fix.
Why now: The reversal shipped July 24, three days after the July 21 restriction it undoes, per Claude Code's changelog.
Claude Code restored nested subagent spawning at depth 3 on July 24, three days after disabling it entirely, per the changelog. It's back on by default for anyone running Claude Code, and it restores an orchestration capability that had been pulled just days earlier.
Depth 3 means a coordinator agent can spawn worker agents that spawn their own verifiers. That shape was impossible under the depth-1 default enforced the week before. It's the difference between one agent doing a task and a small team checking each other's work.
v2.1.217 shipped July 21 and disabled nested subagent spawning by default after unbounded fan-out problems, per the changelog. v2.1.219 shipped July 24 and reversed it, setting the default spawn depth back to 3.
The same environment variable, CLAUDE_CODE_MAX_SUBAGENT_SPAWN_DEPTH, now works in reverse: it's the manual switch to force depth back down to 1. If you set it last week to route around the July 21 restriction, it's still set. It's now suppressing a capability you probably want back.
The changelog doesn't say whether the reversal applies retroactively to users who never touched the setting, only that the default changed.
Two contradictory defaults in three days show Claude Code's team is still tuning fan-out limits by feel. There's no settled model yet for safe spawn depth. Watch whether the next release holds depth 3 or reverses again. Also watch whether anyone who set the env var workaround before July 24 still has it silently overriding a default they'd want back.
Each link below shares sources, entities, or timing with this story.
Two changes shipped within a day of each other that compound in exactly the wrong direction, and nobody connected them. Anthropic's Opus 5 prompting guide documents that the model "delegates to subagents more readily than prior models," and is blunt about the consequence: it "...
One line in the v2.1.215 changelog, July 19: Claude Code no longer invokes the /verify and /code-review skills on its own. You call them explicitly now (changelog). If your workflow assumed a review pass fired at the end of a task, it doesn't anymore, and nothing told you. Cod...
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...
It scans prompts and tool descriptions for patterns written for older model generations. This targets a real invisible tax: system prompts and MCP tool descriptions accumulate verbose few-shot blocks, defensive formatting instructions, and reasoning nudges that earned their ke...
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...
makecindy/cindy (created July 22, 147 forks, v0.1.22 on July 30) is an Apache-2.0 desktop/mobile client treating Claude Code and Codex as interchangeable harnesses: models and harnesses mix freely and can switch mid-task while workspace, memory, skills, and tools stay continuo...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.