HighVulnerabilityLLM-specific
CVE-2026-75858: CodeWhale rlm_eval remote code execution via prompt injection
- Source
- NVD(opens in a new tab)
- Identifier
- CVE-2026-75858
- Published
- Record updated
- Affected
- deepseek-tui >= 0.8.33, <= 0.8.41
- deepseek-tui >= 0.8.33, < 0.8.41, fixed in 0.8.41, released July 27, 2026, 21 days before disclosure
- codewhale-tui >= 0.8.41, < 0.8.64, fixed in 0.8.64, released June 22, 2026, 57 days before disclosure
- and 1 more
- Known exploitation
- Not listed in the CISA Known Exploited Vulnerabilities catalog at the last check.
- EPSS
- 0.4%
Summary
CodeWhale (packages codewhale and codewhale-tui), versions >= 0.8.41 and < 0.8.64, contains a remote code execution flaw in the rlm_eval tool. Its approval_requirement() returns ApprovalRequirement::Auto, so the engine never prompts and runs model-supplied Python in a python3 interpreter without consulting --approval-policy. An attacker can trigger this through prompt injection in untrusted content the agent reads, such as a web page, fetched URL, repository file or MCP tool result, with the companion rlm_open tool able to stage that content. Code runs at the user's privilege level.
Mitigation
Fixed in 0.8.64.
Topics
Related items
- HighCVE-2026-129995: mcp-ssh-manager shell command injection through backup tool argumentsSimilar attack · NVD/CVE Database
- MediumCVE-2026-77332: memtomem-stm envelope accepts Markdown and HTML from memory fieldsSimilar attack · NVD/CVE Database
- InfoMapping the AI threat landscape: MITRE ATLAS comes to Elastic SecuritySimilar attack · Elastic Security Labs
- HighCVE-2026-133631: FlowIntel note variable resolution exposes private case dataSimilar attack · NVD/CVE Database
- InfoAdversarial prompt detection via contrastive dual-encoder transformer for zero-day jailbreak defense based on AdverSentinel-LLM modelSimilar attack · OpenAlex (peer-reviewed AI security)