External risk intelligence

Agent Development Kit Continuation Forgery Vulnerability

CVE advisorySeverity: CRITICAL (CVSS 9.3)

CVE-2026-18236

The vulnerability resides within an Agent Development Kit (ADK), which is a framework or library used by developers to build software agents. Development kits and their internal session management logic are typically used in build-time or developer-specific environments and are not themselves deployed as public-internet-facing services or gateways.

Halo Surface Signal: 1 out of 5 — much less likely to be public-facing.

External exposure likelihood

Horizon Alert

Summary of the vulnerability and why it matters

A vulnerability in the Agent Development Kit (ADK) could allow unauthorized execution of tools by forging confirmation responses, stemming from insufficient verification of tool registration, confirmation requirements, and argument matching.

  • Forged tool confirmations bypass security checks.
  • Matters if ADK is used in sensitive tool workflows.
  • Confirm relevance and exposure within development environments.

Attack Path

How an attacker could exploit the issue

An attacker could exploit this vulnerability by manipulating the session history to forge tool confirmation responses. This bypasses security checks that ensure the correct tool is registered, requires confirmation, and matches the original call, potentially leading to unauthorized tool execution.

  • No authentication or network access required.
  • Manipulating session history to forge confirmations.
  • Unauthorized tool execution.

Live Threat

Current exploitation, exposure, and threat context

A vulnerability in the Agent Development Kit (ADK) could allow an attacker to execute unauthorized tools by forging a tool confirmation response. This could occur if an attacker can manipulate or inject events into the session history and the ADK fails to properly verify the target tool or its confirmation requirements.

  • Tool execution could be compromised.
  • Events can be manipulated in session history.
  • Unauthorized tools may be executed.

Operational Fix

Recommended remediation, mitigation, and detection steps

The Agent Development Kit (ADK) is a development framework, and vulnerabilities within it typically fall under the responsibility of development or platform teams responsible for managing these tools. The initial step is to confirm if the ADK is used in your environment, identify the accountable development team, and assess the risk based on its integration and reach.

  • Development or Platform teams should own.
  • Verify ADK usage and exposure first.
  • Plan remediation based on identified risk.

Supplementary metadata

Validate whether this threat affects your internet-facing exposure.

Halo Threat Intelligence helps prioritize remediation with Halo Surface Signal and H/A/L/O context. Start exposure validation with a free external attack surface trial.

Frequently asked questions

What is the Agent Development Kit?

The Agent Development Kit (ADK) is a software framework or library provided to developers to build autonomous software agents. It simplifies how these agents interact with external tools and manage session state, essentially acting as the plumbing for agent-to-tool communication. Because it is a library rather than a finished application, it is typically utilized during the software development lifecycle to construct and test agent logic.

What is CWE-863 in the context of CVE-2026-18236?

CWE-863 refers to Incorrect Authorization. In this CVE, the ADK fails to properly check if an agent has permission to execute a specific tool or if that tool actually requires user consent. Because the framework does not validate that the confirmation response matches the original request, it allows an attacker to bypass authorization steps and trick the agent into performing actions that were never genuinely requested or approved.

How can an attacker trigger this forgery?

An attacker must be able to inject or manipulate the session history events that the ADK relies on to manage tool calls. By inserting a fake confirmation response into this stream, they can trick the agent into proceeding. This vulnerability does not trigger if the session history is immutable, protected from external tampering, or if the agent implementation performs manual, out-of-band verification of every tool execution request.

Why is Halo Surface Signal labeling this unlikely?

Halo Surface Signal assigns a 'Very unlikely' label because the ADK is a development framework rather than a service exposed to the public internet. Since these libraries are primarily used within build-time or developer environments, they rarely exist on the network perimeter. The risk is generally contained to internal development platforms rather than public-facing gateways or production user interfaces.

Do I need to take action if I use the ADK?

Yes. Start by identifying which of your internal applications or agent projects rely on the ADK library. Coordinate with the development or platform teams managing those projects to assess where tool confirmation is handled. Since this involves internal logic, your priority is to verify if your specific implementation is reachable by untrusted actors and to plan for an update to the secure version of the library once available.

References