Research brief / AI systems
Browser Agent Consultation Prompt Pack
Reusable consultation patterns for browser agents that separate observation, judgment, and action.
The research question
How can an agent preserve the meaning of a task as it moves between models, tools, and context windows?
Inside the work
Make the source intent explicit, separate a proposed action from its authority, and keep a recoverable record of the result. Compare behavior against the original task instead of judging fluency alone.
Research into compiling human intent into a structure an agent can actually execute. The ALX project offers a concrete way into this field of research.
Explore ALXWhat this brief establishes
This is an overview of the program and its place in the research library. It is not a full paper, a dataset, or evidence of a measured outcome. Where a related project has a tested result or an open limitation, its project page names that boundary.
In the same field
ALX Semantic Compiler
Source-preserving semantic compilation, execution contracts, and bounded repair with explicit evaluation limits.
Protocol Compiler Framework
A repeatable method for compiling research and operating rules into agent-usable protocols.
ALX LoRA Evaluation Track
Evaluation plans for whether specialized training can preserve ALX structure and reversibility.