by Kastro
Paper/Archive/CVE-2026-45374

CVE-2026-45374

Hmbown CodeWhale — CodeWhale: task_create Insecure Defaults Enable RCE via Prompt Injection in Project Files

EPSS 0.4%public exploitCVSS 9.6
critical.is priorityof 100
45
CVSS 9.6 (CNA)9.6 / 10 × 3028.8Not in CISA KEV+300.0EPSS percentile 32.9.329 × 206.6Exploit: SSVC poc+1010.0SSVC exploitation: poc+100.0
Description · from the CVE record

CodeWhale is a DeepSeek + MiMo coding agent in terminal. Prior to 0.8.26, the task_create tool spawns durable sub-agents that inherit two insecure defaults, allow_shell defaults to true (config.rs:1499: self.allow_shell.unwrap_or(true)) and auto_approve defaults to true (task_manager.rs:297: auto_approve: Some(true)). When a user approves a task_create call (which requires ApprovalRequirement::Required), they approve what appears to be a benign work prompt. However, the spawned sub-agent silently receives unrestricted, unapproved shell access. This vulnerability is fixed in 0.8.26.

Affected

ProductAffectedFixed inSource
CodeWhale< 0.8.26–CNA

Scores by source

GitHub_M (CNA)9.6CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:H
CISA ADPSSVCExploitation: poc · Automatable: no · Technical impact: total

References

exploitgithub.com/Hmbown/CodeWhale/security/advisories/GHSA-72w5-pf8h-xfp4refgithub.com/Hmbown/DeepSeek-TUI/releases/tag/v0.8.26