# CVE-2026-51866: DB-GPT code execution in skill manager

Source: https://junglewise.ai/threats/cve-2026-51866-in-db-gpt-0-7-5-and-0-8-0-a-skill-uploaded-through-the-real-api  
Publisher: Junglewise Threat Intelligence, https://junglewise.ai/threats  
Published: 2026-09-30. Updated: 2026-09-30  
License: CC BY 4.0 with a link back

CVE-2026-51866 · Severity: info · Published 2026-09-30

DB-GPT is a framework for building AI agents with plugin capabilities. An attacker can upload a malicious skill through the public /api/v1/skills/upload endpoint and then trigger its execution on the server through the /api/v1/chat/react-agent workflow, gaining arbitrary code execution on the host system.

Technologies: [Eosphoros-Ai DB-GPT](https://junglewise.ai/threats/technologies/db-gpt). Vendors: [Eosphoros-Ai](https://junglewise.ai/threats/vendors/eosphoros-ai).

## Quick answers

**Is CVE-2026-51866 exploited in the wild?** No. Junglewise Threat Intelligence has no report of CVE-2026-51866 being exploited in the wild.

**How severe is CVE-2026-51866?** Informational severity, published 30 September 2026.

**Which products and versions are affected?** Eosphoros-Ai DB-GPT 0.7.5, 0.8.0.

**Is there a fix for CVE-2026-51866?** No fixed version is recorded yet. Check the vendor's advisory in the references below. If you run it, Junglewise drafts a mitigation that fits your setup and applies it once you approve, with senior security engineers behind it.

**How is CVE-2026-51866 attacked?** It is attacked over the network, in a software library.

**Am I affected by CVE-2026-51866?** Check whether you run Eosphoros-Ai DB-GPT at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

## Technical details

The vulnerability exists in SkillManager.execute_skill_script_file, which executes uploaded Python scripts via asyncio.create_subprocess_exec without adequate trust separation. An attacker uploads a skill containing arbitrary Python code, and when the chat agent is directed to use that skill, the code executes with server privileges. The attack vector is network-based and requires interaction with the chat agent to select the uploaded skill.

## Affected products

- eosphoros-ai DB-GPT 0.7.5, 0.8.0

## Timeline

- 2026-05-07: disclosed
- 2026-09-30: disclosed: CVE-2026-51866 published

## References

- [https://gist.github.com/Ro1ME/f8f5b730a4000684d3a92a67f82ee03c](https://gist.github.com/Ro1ME/f8f5b730a4000684d3a92a67f82ee03c)
- [https://github.com/eosphoros-ai/DB-GPT/issues/3047](https://github.com/eosphoros-ai/DB-GPT/issues/3047)

## Related threats

- [eosphoros-ai DB-GPT remote code execution in HuggingFace adapter](https://junglewise.ai/threats/cve-2026-123019-a-weakness-has-been-identified-in-eosphoros-ai-db-gpt-up-to-0-8)
- [eosphoros-ai DB-GPT code injection in CodeAction](https://junglewise.ai/threats/cve-2026-123018-a-security-flaw-has-been-discovered-in-eosphoros-ai-db-gpt-up-to)
- [DB-GPT directory traversal in use_example_file](https://junglewise.ai/threats/cve-2026-51863-db-gpt-v0-7-5-and-0-8-0-contains-directory-traversal-in-use)
- [DB-GPT hardcoded credentials in default Docker configuration](https://junglewise.ai/threats/cve-2026-77933-db-gpt-through-0-8-2-contains-a-hardcoded-credentials)
- [DB-GPT unauthenticated remote code execution](https://junglewise.ai/threats/cve-2026-77932-db-gpt-through-0-8-2-contains-an-unauthenticated-remote-code)