CVE-2026-61732: Decepticon: Role-boundary forgery via ChatML special-token literals in web crawl output composed into LLM context

Should you act?

Critical and remotely reachable

No known exploitation yet, but it is network-reachable without credentials. This signal is computed from CISA KEV, EPSS and CVSS data; it is not an editorial verdict.

• Not in CISA KEV, but CISA’s SSVC assessment records a public proof-of-concept, so attack code exists.
• EPSS 1.2%, percentile 68: FIRST’s model estimates a 1.2% chance of exploitation activity in the next 30 days, higher than 68% of all scored CVEs (moderate). EPSS is a probability, not a verdict.
• Remotely reachable with no login and no user interaction. If the vulnerable service is exposed to the internet, prioritize it.
• Only applies if your dependency tree includes decepticon-core (PyPI) <= 1.1.16, fixed in 1.1.17; decepticon (PyPI) <= 1.1.16, fixed in 1.1.17; decepticon-sdk (PyPI) <= 1.1.16, fixed in 1.1.17. Check your lockfile, including transitive dependencies.

What it is

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.

From the GitHub advisory

Summary

Decepticon wraps 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.

The vast majority of open-source and self-deployed model providers do not filter special-token literals. vLLM explicitly declined to fix this issue on 2026-04-21, closing it as "out of scope for the inference layer." Fix responsibility therefore falls squarely on the Agent application layer. OpenClaw completed an analogous fix on 2026-04-22 via commit 2514746b3261 (~30 lines, sanitizer applied just before tool-output wrapping), demonstrating the feasibility of application-layer mitigation.

Applicability

Confirmed vulnerable when Decepticon is configured with a BYOK OpenAI-compatible backend whose tokenizer preserves special-token IDs — vLLM / SGLang / TGI confirmed upstream.

Weakness type: CWE-74.

Product: decepticon-core (PyPI) · Assigned by: GitHub_M · Published: 2026-09-24

Affected and fixed versions

Severity scores

• Attack vector: network (remotely reachable)
• Attack complexity: low
• Privileges required: none (no login)
• User interaction: none
• Scope: changed (can affect other components)

CVSS measures how bad a flaw is if exploited, not how likely exploitation is. Read it next to the exploitation data below.

Exploitation and urgency

• CISA KEV: not listed (no confirmed in-the-wild exploitation known to CISA).
• CISA SSVC assessment: exploitation poc, automatable yes, technical impact total.
• FIRST EPSS (model of 2026-10-21): 1.22% probability of exploitation activity in the next 30 days; percentile 68.

Timeline

• 2026-09-24 — CVE record last updated
• 2026-09-24 — CVE record published by GitHub_M
• 2026-09-24 — GitHub-reviewed advisory GHSA-g5f9-3xfg-p9mf published (critical)

Sources and references

Original sources first. Version and fix details should always be confirmed against the vendor or project advisory.

Vulnerability records

1. GitHub Advisory GHSA-g5f9-3xfg-p9mf github.com
2. CVE.org record CVE-2026-61732 cve.org
3. NVD entry CVE-2026-61732 nvd.nist.gov

Other references from the CVE record and advisory

1. github.com github.com
2. github.com github.com
3. github.com github.com
4. github.com github.com

Critical PyPI advisories · Today’s Security Digest