Prompts
The task each variant gives the agent.
prompts defines the tasks agents try to complete. Each resolved variant receives one final prompt. Write it as a realistic, self-contained job. Language that tells the agent it is being evaluated changes how it works; see Experiment design.
Accepted shapes:
# String form: one prompt.
prompts: "Build the report."
# List of strings: multiple prompts.
prompts:
- "Build the report."
- "Build the report and explain your steps."
# Object form: list of named prompts.
prompts:
- id: detailed
prompt: "Build the report and explain your steps."Prompt object
A prompt object gives a task stable metadata, so Results can filter and group by the prompt that produced each run.
| Field | Required | Type / values | Notes |
|---|---|---|---|
id | Yes | Kebab-case string | Stable prompt id. Recorded for filtering in Results. |
prompt | Yes | String | Task text given to the agent. Describe a realistic job; do not say the agent is being evaluated. |
description | No | String | Human-readable notes. |
tags | No | String list | Free-form labels. |
Bare prompt strings get positional ids: p0, p1, and so on.