# vLLM before 0.28.0 contains a remote code execution vulnerability in the LlavaOnevision2 processor loader that ignores the trust_remote_code parameter when loading remote processor classes. Attacke... Advisory record · `GHSA-3c86-2m5g-59q7` - **Identifiers:** `GHSA-3c86-2m5g-59q7` · `CVE-2026-90553` · `PYSEC-2026-3985` - **State:** Confirmed - **Severity:** CVSS 8.5 High - **Critical:** No - **Weakness (CWE):** [CWE-94](https://cwe.mitre.org/data/definitions/94.html) - **Tags:** [Code injection](https://agentprime.io/advisories/tag/code-injection.md) - **Published:** 12 Sept 2026 - **Updated:** 17 Sept 2026 - **Reviewed by a person:** No [Published by the maintainers](https://github.com/vllm-project/vllm/security/advisories/GHSA-3c86-2m5g-59q7) · not yet reviewed by GitHub. The vllm-project/vllm maintainers published this as a GitHub repository advisory; GitHub’s reviewed advisory replaces it here when it arrives. Published by the vllm-project/vllm maintainers at 12 Sept 2026, 13:16 UTC. agentprime has not analysed this yet, so whether it affects any app is not known. ## Affected packages - `vllm` (PyPI): affected `<0.28.0`; fixed in `0.28.0` ## Machine-written summary Summary written by Claude Haiku from the source advisory. Not reviewed by a person. > vLLM versions before 0.28.0 are vulnerable to remote code execution in the LlavaOnevision2 processor loader, which loads remote processor classes without honoring the trust_remote_code setting. An attacker who can supply a malicious model can run arbitrary code with the privileges of the vLLM process, even when trust_remote_code is False. Upgrade to vLLM 0.28.0 or later. Written 1 Nov 2026. ## Not analysed yet We haven’t published our analysis of this Advisory yet. Once a person has reviewed it, this page will say what’s vulnerable, what has to be true for your app to be affected, and how to check. ## Severity and exploitation - 8.5 High, CVSS 4.0: `CVSS:4.0/AV:L/AC:L/AT:N/PR:N/UI:P/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N`, from CNA (VulnCheck) - **CISA KEV:** Not listed. - **EPSS:** 0.0031, the 23rd percentile, on 12 Nov 2026: FIRST’s estimate of the chance of exploitation in the next 30 days. ## Timeline - 28 Aug 2026, 18:05 UTC: [Published by the maintainers as a GitHub repository advisory](https://github.com/vllm-project/vllm/security/advisories/GHSA-3c86-2m5g-59q7) - 31 Oct 2026, 14:10 UTC: First seen by agentprime - 31 Oct 2026, 14:10 UTC: [Confirmed by OSV](https://osv.dev/vulnerability/PYSEC-2026-3985) ## Sources - [GitHub Advisory Database: GHSA-3c86-2m5g-59q7](https://github.com/advisories/GHSA-3c86-2m5g-59q7) - [OSV: GHSA-3c86-2m5g-59q7](https://osv.dev/vulnerability/GHSA-3c86-2m5g-59q7) - - Facts on this page come from the GitHub Advisory Database (CC-BY 4.0), OSV and the other sources listed, as published. We don’t reproduce full advisory descriptions, which can contain exploit code. ## Check whether this affects you A Match means an installed version is in the affected range above. Most don’t affect the app; agentprime hasn’t analysed this one yet. - With the agentprime MCP server (), ask your agent to check the project: `check_lockfile` sends a lockfile, `check_packages` a list of installed packages, and both return the Matches and the upgrades that clear them. `get_advisory` returns this record. Setup: - In a browser: reads your lockfile in the browser, never uploading it. - Whether it touches your code, with file and line: the Claude Code plugin, - As JSON: , or `POST https://api.agentprime.io/v1/match` with your packages () --- The web page: . Everything for agents: .