Watch card · compiled automatically, not analysed
CVE-2026-96620
Claude Agent SDK: Argument Injection via resume Option Allows Arbitrary Command Execution
AI summaryClaude Agent SDK (Python) before 0.2.121 doesn't validate the resume option, so an attacker could inject an MCP server config and run commands.
Our research pipeline compiles this page from public data. It contains no human judgement from PlainCVE; rely on the original advisories for facts.
Our status
Written up — We have written this up; read the full analysis. Read the write-up →
Key numbers
- CVSS
- 9.2 v4.0
- CISA KEV
- Not listed
- EPSS
- 0.58%
- Momentum
- —
Timeline
- 2026-10-16DisclosedNVD
How much time similar vulnerabilities gave
This item is not mapped to a category yet, so here are the overall numbers: of 390 vulnerabilities disclosed since 2025 and added to KEV, the median time from disclosure to KEV was 5 days; 54% were added within 7 days.
This one: not in KEV so far (that does not mean it is not exploited).
Counts only vulnerabilities CISA has confirmed as exploited; disclosure is the NVD publication date, which can differ from the vendor advisory.
Affected packages (OSV)
| Ecosystem | Package | From | Fixed in |
|---|---|---|---|
| PyPI | claude-agent-sdk | all earlier versions | 0.2.121 |
Primary sources
Watch cards update with This week every 3 hours. Once we write this up, this URL will point to the write-up.