Fetching from the wire…
Public story · 2026-07-19 · high
The same July release tightens permission checks and adds prompts before docker commands, per Claude Code's docs.
Why now: July's release notes are the newest documentation of unattended Claude Code behavior, as of July 19.
Claude Code shipped EndConversation in July's release, a tool that lets an agent declare it's finished, per the product's docs.
Both are prerequisites for running Claude Code as a cron job instead of a supervised session. A hung or looping process running unattended used to just sit there burning compute until someone happened to check on it.
Before EndConversation, a finished agent had no clean way to say so, per the docs. It would idle, or loop back into work it had already done, wasting the run.
The same release added progress heartbeats for long-running tasks. A supervising process used to have to guess whether a task was still working or just hung; heartbeats give it a signal instead.
July's release also tightens permission checks and adds prompts before docker commands, according to the docs. It doesn't say whether those prompts fire on workflows that already have docker approved, or on every invocation.
Each link below shares sources, entities, or timing with this story.
The June 25 release adds autoMode.classifyAllShell to route every Bash and PowerShell command through the auto-mode classifier, surfaces denial reasons in the transcript, and emits a claude_code.assistant_response OTel event carrying the model's actual response text. (Claude C...
v2.1.206 (July 10) puts a tabbed Browser panel behind Cmd+Shift+B that Claude drives to read docs, inspect mockups, and fill forms, running in an isolated profile separate from your real credentials, with a domain allowlist and per-site permission cards (Claude Code Docs). Thi...
Three frontier models shipped in a single week this month, and teams with a standing eval harness had a routing decision in hours. Anthropic's own agent-eval guidance says 20-50 tasks drawn from your real usage and real failures is enough to detect issues (DeepEval). DeepEval...
Two days from now, on August 14, auto mode becomes the default permission mode for new Pro, Max, and Team sessions (Claude Code Docs, Week 32). Not opt-in. Default. Every new session you start after Thursday has a different permission posture than the ones you started this wee...
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...
Two researchers in my set surfaced this independently, which is usually a sign it matters. Claude Code 2.1.166, first seen June 6, introduces a fallback-models setting: configure up to three models tried in order when the primary is overloaded or unavailable. It also adds glob...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.