Claude Agent SDK for Python: command injection via resume option

published 2026-10-16

In claude-agent-sdk below 0.2.121 the resume option allows argument injection, so crafted input leads to arbitrary command execution. It is rated critical.

Fixed in: 0.2.121

CVE: CVE-2026-96620

Source: GitHub Advisory

Am I affected?

You are affected if the installed version is older than 0.2.121. Check the version you have and update to 0.2.121 or later. The fixed version is verified against a vendor or advisory source.

Get the fix

FixLane has a reviewed fix procedure for this incident: a read-only check first, then the fix. Ask the Telegram bot, or connect an AI agent to the FixLane MCP server and call get_fix_prompt with incident id 1801 and your operating system.