Claude Discovery

← All discoveries

Close-up of a smartphone wrapped in a chain with a padlock, symbolizing strong security.
Photo by Towfiqu barbhuiya on Pexels
tool

Claude Code Patches a Bash Permission Bypass in 2.1.223

2026-08-07 ยท source:

Claude Code 2.1.223 fixes a Bash permission bypass where a crafted command could hide part of itself from the permission check, and adds owner wildcard entries for marketplace allow and block lists.

What it is

Claude Code 2.1.223 is a security-and-features patch release for Anthropic's coding agent CLI.

What it does

The release closes a Bash permission bypass where a specially crafted command could conceal parts of itself from the permission check, letting disallowed commands slip through. It also adds owner wildcard entries (`"owner/*"`) to the `strictKnownMarketplaces` and `blockedMarketplaces` managed settings, so admins can allow or block every marketplace repo under a GitHub org in one rule instead of listing them individually. A new warning fires when a workflow agent, forked skill, slash command, or resumed background agent requests a subagent model that's restricted, so the parent model silently taking over doesn't go unnoticed.

Why it matters

Bash permission-check bypasses are exactly the kind of gap that turns a sandboxed coding agent into an unsandboxed one, echoing the worktree-isolation fix Anthropic shipped just a version earlier. The org-wide marketplace wildcard is a small but real convenience for enterprises managing plugin sources at scale.

How to use it

Update to Claude Code 2.1.223 or later, and set `"owner/*"` entries in `strictKnownMarketplaces` or `blockedMarketplaces` managed settings to cover an entire GitHub org.

Go to source →
claude-codesecurityrelease-notespermissions