ACDL: A New Language for Describing LLM Agent Contexts
A new language called the Agentic Context Description Language (ACDL) has been introduced to precisely specify the structure and dynamics of LLM input contexts in agentic systems. As large language models are increasingly used within larger systems that make sequences of LLM calls, the design of encoded information and its structure—known as context engineering—plays a central role in system quality. Currently, context construction is conveyed through informal prose, ad hoc diagrams, or code inspection, which lack precision in capturing prompt evolution across interaction steps or differences between context representation strategies. ACDL aims to remedy this by providing a standardized way to communicate how LLM context is composed and evolves over time. The language is detailed in a paper on arXiv (2605.01920v1).
Key facts
- ACDL stands for Agentic Context Description Language.
- It is designed to specify LLM input context structure and dynamics.
- LLM agents make sequences of LLM calls with instructions, observations, and history.
- Context engineering is critical for system quality.
- Current methods use informal prose, diagrams, or code.
- ACDL provides a precise standard for context description.
- The paper is on arXiv with ID 2605.01920v1.
- The language addresses prompt evolution and strategy comparison.
Entities
Institutions
- arXiv