AI Agent Tools
Sub-technique of AML.T0017 Develop Capabilities
Description
Adversaries may develop or materially adapt tools, integrations, or tool servers designed to extend the capabilities of an AI agent. These capabilities may allow an agent to interact with operating systems, browsers, networks, cloud services, data stores, software repositories, identity systems, or other external resources.
AI agent tools may be implemented as Model Context Protocol servers, plugins, skills, connectors, function libraries, computer-use adapters, execution brokers, remote APIs, or similar model-callable interfaces. Development may include creating executable functionality, model-visible tool descriptions, procedural instructions, input schemas, authentication methods, permission handling, or packaging needed to make a capability available to an agent.
- ATLAS ID
- AML.T0017.002
- Maturity
- RealizedUsed by real adversaries against AI systems in the wild.
- Platforms
- Predictive AIGenerative AIAgentic AIEnterprise
Mitigations
ATLAS 2026.09 lists no mitigation for this technique; see the parent technique.
Case studies that use it
Related ATT&CK technique
ATLAS references this ATT&CK technique on the parent technique.