CVE-2026-59820: LiteLLM Arbitrary File Write via Skills Archive

# CVE-2026-59820: LiteLLM Arbitrary File Write via Skills Archive

https://gcdradar.vercel.app/incidents/cve-2026-59820-litellm-arbitrary-file-write-via-skills-archi · AIR-2026-0833 · closed 2026-07-08

- Lane: Patch now
- Severity: 2 (Disclosed vulnerability)
- Fix status: Fix status unknown
- First reported: 2026-07-08 (happened 2026-07-22)
- AI role: Attack on AI
- Products: LiteLLM
- CVE ids: CVE-2026-59820
- CVSS: 6.1

## Summary

LiteLLM skills archive extraction failed to properly validate file paths from uploaded ZIP archives, leading to path traversal. Authenticated users could exploit this to write arbitrary files and potentially achieve code execution.

## What to do

No action recorded; read the sources below and the vendor's advisory.

## Latest updates

- 2026-07-22: CVE-2026-59820: LiteLLM: Arbitrary file write via path traversal in Skills archive extraction (https://github.com/advisories/GHSA-5jmr-gcrj-2c9q)
- 2026-07-08: CVE-2026-59820: LiteLLM is a proxy server (AI Gateway) to call LLM APIs in OpenAI (or native) format. (https://nvd.nist.gov/vuln/detail/CVE-2026-59820)

## Sources

- 2026-07-08 NVD: CVE-2026-59820: LiteLLM is a proxy server (AI Gateway) to call LLM APIs in OpenAI (or native) format. https://nvd.nist.gov/vuln/detail/CVE-2026-59820
- 2026-07-22 GitHub Advisory Database: CVE-2026-59820: LiteLLM: Arbitrary file write via path traversal in Skills archive extraction https://github.com/advisories/GHSA-5jmr-gcrj-2c9q

Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.