CVE-2026-96620

Status: Awaiting AnalysisCritical (9.2)—

Severity9.2Critical · CVSS 4.0

Exploitation probability0.58 %higher than 47% of CVEs

Active exploitationNonot in the CISA catalog

Public exploitNot knownno public code found

Claude Agent SDK is a Python SDK for Claude Code. Prior to 0.2.121, the resume option is passed to the Claude CLI as a separate token without validation, so a dash-leading value is parsed as an independent CLI flag. An attacker can inject the --mcp-config flag to define a stdio MCP server, causing arbitrary command execution at CLI startup outside the SDK permission controls. The issue is exploitable when an application passes untrusted input as the session ID to resume. This issue is fixed in version 0.2.121.

CVSS

Base score: 9.2
Version: 4.0
Vector: CVSS:4.0/AV:N/AC:L/AT:P/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X

Exploitation probability (EPSS)

Probability of exploitation in the next 30 days: 0.58 %
Percentile among all scored CVEs: 47
Score date: 10/20/2026

EPSS (Exploit Prediction Scoring System, FIRST) estimates how likely a vulnerability is to be exploited in the wild within 30 days. It complements CVSS (impact) and CISA KEV (confirmed exploitation).

ATT&CK techniques

How this vulnerability is exploited and what the attacker gains, in MITRE ATT&CK terms.

ExploitationT1190 Exploit Public-Facing Applicationinitial access90 %
Primary impactT1059 Command and Scripting Interpreterexecution90 %

Vector AV:N/PR:N/UI:N y la descripción: un valor resume con guion inyecta --mcp-config y ejecuta comandos arbitrarios al arrancar la CLI.

Inferred by our analysis agent from the official description, CVSS vector and CWE, and checked by a supervisor. May contain errors.

ATT&CK mitigations that cover these techniques

Privileged Account Management Vulnerability Scanning Restrict Web-Based Content Network Segmentation Limit Software Installation Limit Access to Resource Over Network

Affected technologies (1)

⚠ AI-inferred from the description — NVD hasn't analyzed this CVE yet, these aren't verified CPEs.

Anthropic — Claude Agent SDK^AI

CWEs

CWE-20, CWE-78, CWE-88

References

Show 4 references

Raw JSON (NVD)

Show

{
  "id": "CVE-2026-96620",
  "cveTags": [],
  "metrics": {
    "cvssMetricV40": [
      {
        "type": "Secondary",
        "source": "security-advisories@github.com",
        "cvssData": {
          "Safety": "NOT_DEFINED",
          "version": "4.0",
          "Recovery": "NOT_DEFINED",
          "baseScore": 9.2,
          "Automatable": "NOT_DEFINED",
          "attackVector": "NETWORK",
          "baseSeverity": "CRITICAL",
          "valueDensity": "NOT_DEFINED",
          "vectorString": "CVSS:4.0/AV:N/AC:L/AT:P/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X",
          "exploitMaturity": "NOT_DEFINED",
          "providerUrgency": "NOT_DEFINED",
          "userInteraction": "NONE",
          "attackComplexity": "LOW",
          "attackRequirements": "PRESENT",
          "privilegesRequired": "NONE",
          "subIntegrityImpact": "NONE",
          "vulnIntegrityImpact": "HIGH",
          "integrityRequirement": "NOT_DEFINED",
          "modifiedAttackVector": "NOT_DEFINED",
          "subAvailabilityImpact": "NONE",
          "vulnAvailabilityImpact": "HIGH",
          "availabilityRequirement": "NOT_DEFINED",
          "modifiedUserInteraction": "NOT_DEFINED",
          "modifiedAttackComplexity": "NOT_DEFINED",
          "subConfidentialityImpact": "NONE",
          "vulnConfidentialityImpact": "HIGH",
          "confidentialityRequirement": "NOT_DEFINED",
          "modifiedAttackRequirements": "NOT_DEFINED",
          "modifiedPrivilegesRequired": "NOT_DEFINED",
          "modifiedSubIntegrityImpact": "NOT_DEFINED",
          "modifiedVulnIntegrityImpact": "NOT_DEFINED",
          "vulnerabilityResponseEffort": "NOT_DEFINED",
          "modifiedSubAvailabilityImpact": "NOT_DEFINED",
          "modifiedVulnAvailabilityImpact": "NOT_DEFINED",
          "modifiedSubConfidentialityImpact": "NOT_DEFINED",
          "modifiedVulnConfidentialityImpact": "NOT_DEFINED"
        }
      }
    ]
  },
  "affected": [
    {
      "source": "security-advisories@github.com",
      "affectedData": [
        {
          "vendor": "anthropics",
          "product": "claude-agent-sdk-python",
          "versions": [
            {
              "status": "affected",
              "version": "< 0.2.121"
            }
          ]
        }
      ]
    }
  ],
  "published": "2026-10-16T20:18:28.293",
  "references": [
    {
      "url": "https://github.com/anthropics/claude-agent-sdk-python/commit/347a1cb88d7652199399db1b7f914e0a459a6662",
      "source": "security-advisories@github.com"
    },
    {
      "url": "https://github.com/anthropics/claude-agent-sdk-python/pull/1123",
      "source": "security-advisories@github.com"
    },
    {
      "url": "https://github.com/anthropics/claude-agent-sdk-python/releases/tag/v0.2.121",
      "source": "security-advisories@github.com"
    },
    {
      "url": "https://github.com/anthropics/claude-agent-sdk-python/security/advisories/GHSA-h4mw-j7qp-8mwm",
      "source": "security-advisories@github.com"
    }
  ],
  "vulnStatus": "Awaiting Analysis",
  "weaknesses": [
    {
      "type": "Primary",
      "source": "security-advisories@github.com",
      "description": [
        {
          "lang": "en",
          "value": "CWE-20"
        },
        {
          "lang": "en",
          "value": "CWE-78"
        },
        {
          "lang": "en",
          "value": "CWE-88"
        }
      ]
    }
  ],
  "descriptions": [
    {
      "lang": "en",
      "value": "Claude Agent SDK is a Python SDK for Claude Code. Prior to 0.2.121, the resume option is passed to the Claude CLI as a separate token without validation, so a dash-leading value is parsed as an independent CLI flag. An attacker can inject the --mcp-config flag to define a stdio MCP server, causing arbitrary command execution at CLI startup outside the SDK permission controls. The issue is exploitable when an application passes untrusted input as the session ID to resume. This issue is fixed in version 0.2.121."
    }
  ],
  "lastModified": "2026-10-19T21:13:39.973",
  "sourceIdentifier": "security-advisories@github.com"
}