alfworld-goal-interpreter

What problem does it solve?

This Skill tackles the complexity of AI agent tasks by breaking down natural language instructions into actionable sub-objectives and identifying necessary objects, making it easier for agents to understand and execute commands.

Core Features & Use Cases

• Natural Language Parsing: Interprets task goals to extract primary targets, reference objects, spatial relationships, and actions.
• Sequential Planning: Generates clear, step-by-step sub-objectives based on parsed information, considering spatial context.
• Object & Action Identification: Maps extracted objects to the agent's available actions and identifies critical steps like opening containers.
• Fallback Logic: Provides strategies for when initial actions fail, referencing known object-location associations and spatial resolution patterns.
• Use Case: When an agent is told to "find the red ball under the table," this skill will parse it to identify "red ball" as the primary target, "table" as the reference object, "under" as the spatial relation, and "find" as the action, then generate a plan to look under the table.

Quick Start

Use the alfworld-goal-interpreter skill to parse the task goal "find the blue cup on the shelf".