{"id":"CVE-2026-61732","published":"2026-09-24T18:17:15.860","lastModified":"2026-09-24T19:17:14.910","description":"Decepticon is an autonomous hacking agent for red teams. Versions prior to 1.1.17 wrap web crawl results — the output of agent reconnaissance against target services — into LLM messages without neutralizing ChatML special-token literals. Under the BYOK (Bring Your Own Key) deployment model, users configure their own LLM credentials to any OpenAI-compatible endpoint. Most open-source and self-deployed model providers (vLLM, SGLang, Ollama, LM Studio, text-generation-webui, etc.) do not filter special-token literals from user content in their default configurations. Those literals are parsed into structural role-boundary token IDs, meaning an attacker string planted in a target web page forges a new operator turn the model treats as authoritative, bypassing Decepticon's agent guardrails and resulting in arbitrary command execution inside the Kali Linux sandbox. Version 1.1.17 patches the issue.","cvssScore":10,"cvssSeverity":"CRITICAL","cvssVector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H","cwes":["CWE-74"],"vendors":[],"products":[],"references":[{"url":"https://github.com/BitterSecurity/Decepticon/commit/79ee2aaf22f4c36a5b1968f6ca3f8086b6e35b67","tags":[]},{"url":"https://github.com/BitterSecurity/Decepticon/security/advisories/GHSA-g5f9-3xfg-p9mf","tags":[]},{"url":"https://github.com/BitterSecurity/Decepticon/security/advisories/GHSA-g5f9-3xfg-p9mf","tags":[]}],"exploitRefs":[{"url":"https://github.com/BitterSecurity/Decepticon/commit/79ee2aaf22f4c36a5b1968f6ca3f8086b6e35b67","tags":[]},{"url":"https://github.com/BitterSecurity/Decepticon/security/advisories/GHSA-g5f9-3xfg-p9mf","tags":[]},{"url":"https://github.com/BitterSecurity/Decepticon/security/advisories/GHSA-g5f9-3xfg-p9mf","tags":[]}],"hasPoc":true,"ai":{"summary":"The flaw allows attackers to bypass security guardrails by injecting special-token literals into web crawl results, leading to arbitrary command execution within the Kali Linux sandbox.","exploitability":"Exploitation is moderately hard as it requires control over the target web page and BYOK deployment model. Precondition is the use of an open-source or self-deployed model provider without special-token filtering.","blast_radius":"If exploited, the impact is high, as it could lead to complete compromise of the Kali Linux sandbox and arbitrary command execution.","remediation":"Upgrade to Decepticon version 1.1.17 or later.","detection":"No reliable host or network indicator is derivable from the published description.","tags":["rce","web","auth-bypass","llm","byok"],"model":"qwen2.5:7b-instruct","analyzedAt":"2026-09-27T08:46:29.270Z"}}