G Connect Applications and AI Agents through Standard Actions

Read recipes, choose extension points and write reviewable Agent tasks. Explicit inputs, outputs and statuses let automation and human operation share capabilities.

Preparation and Outcomes

Start with B1's data mapping, C2's objective interpretation and F1's component responsibilities. G1 is a recipe-reading lesson, G2 assumes Python development experience, and G3 starts with a natural-language task.

Discuss the excerpts and prompts offline. Execution uses the complete release examples, dependencies and valid authorization; extension work also needs an editable calling layer and a test environment. Outcomes are a configuration diff, an extension design with tests, and an Agent task with review records.

1. Read and Modify a Capability Orchestration Recipe
2. Extend a Capability with a Clear Calling Interface
3. Write an AI Agent Task That Can Be Reviewed

Read the three lessons in order. Attempt each exercise before checking the suggested answer, and use the worksheets to record conditions, results and open questions.

Further Reading

Standard Actions and Orchestration · Actions, Recipes and Execution · Developer Guides · Runtime Status and Limits