CVE-2026-96620 anthropics CVE debrief

Who should care

Defenders responsible for applications using the Claude Agent SDK for Python should assess exposure and prioritize updates to mitigate this vulnerability. This includes reviewing application code that passes untrusted input as the session ID to resume and verifying the SDK's permission controls are not bypassed.

Why it matters

The Claude Agent SDK for Python is vulnerable to argument injection, allowing arbitrary command execution. Defenders must update to the latest version and assess exposure in applications that pass untrusted input as the session ID to resume.

• Defenders must update to the latest version of the claude-agent-sdk package to prevent arbitrary command execution.
• Applications passing untrusted input as the session ID to resume are at risk of exploitation.
• The SDK's permission controls can be bypassed, leading to potential security breaches.
• Verification of the SDK's configuration and input validation is necessary to ensure security.

Technical summary

The Claude Agent SDK for Python passes the `resume` option to the Claude CLI without validation, allowing a value beginning with `-` to be interpreted as a separate CLI flag. An attacker who controlled the `resume` value could inject an `--mcp-config` flag defining a stdio MCP server, resulting in arbitrary command execution at CLI startup that bypassed the SDK's permission controls. This issue affects applications that pass untrusted input as the session ID to resume, and defenders should prioritize updating to the latest version of the claude-agent-sdk package to mitigate this vulnerability.

Defensive priority

Defenders should prioritize updating to the latest version of the claude-agent-sdk package to mitigate this vulnerability. Additionally, they should assess exposure in applications that pass untrusted input as the session ID to resume and verify the SDK's permission controls are not bypassed.

Recommended defensive actions

• Update to the latest version of the claude-agent-sdk package
• Assess exposure in applications that pass untrusted input as the session ID to resume
• Verify the SDK's permission controls are not bypassed
• Review compensating controls for exposed systems while remediation is scheduled and verified
• Check relevant monitoring, detection, and logs for exposed assets that need extra review
• Track exceptions, retest remediated assets, and close the item only after evidence is documented
• Confirm whether affected product deployments exist in managed environments and assign an owner for follow-up

Evidence notes

The vulnerability was reported by hackerone.com/moamenmahmood and is documented in the Claude Agent SDK's security advisory. The issue is tracked as GHSA-h4mw-j7qp-8mwm. Evidence of the vulnerability's existence and impact is limited, and defenders should verify the SDK's configuration and input validation to ensure security. The vulnerability affects applications that pass untrusted input as the session ID to resume, and defenders should assess exposure and prioritize updates to mitigate this vulnerability.

Sources and references

Methodology and review provenance

AI-assisted synthesis based on stored public vulnerability evidence. System validation, approval state, and publication status do not by themselves establish human review of this revision. PatchSiren helps prioritize defensive review and does not prove exposure or remediation on any system.