Choose a workflow

Choose the question that matches the decision in front of you. Each guide gives your agent a bounded prompt, a spend boundary, and a report-back contract.

​

Is my AGENTS.md change helping?

Compare one instruction-file lever on the same retained tasks. Read the instruction-file A/B guide.

​

Improve my AGENTS.md

Run a bounded native optimize loop with explicit stop rules. Read the iterative improvement guide.

​

Is this skill helping?

Choose the correct absent or committed baseline before comparing behavior. Read the skill evaluation guide.

​

Which model or reasoning effort should I use?

Smoke first, then compare one new arm against the same frozen reference. Read the model comparison guide.

​

What happens next

Use the prompt cookbook for session-ready wording. Your agent reads the Trial Result and reports the decision, evidence posture, limits, spend, and one next action.