> ## Documentation Index
> Fetch the complete documentation index at: https://icrl.dev/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Prompt Templates

> Design prompts for planning, reasoning, and action generation

## Required Templates

Python and TypeScript `Agent` both require:

* `planPrompt`
* `reasonPrompt`
* `actPrompt`

## Supported Placeholders

* `{goal}`
* `{examples}`
* `{plan}`
* `{observation}`
* `{reasoning}`
* `{history}`

## Template Guidance

* Keep `plan` concise and structured.
* Keep `reason` focused on current observation and progress.
* Keep `act` strict: ask for a single action/command.

## Minimal Set

```text theme={null}
PLAN: Goal: {goal}\nExamples:\n{examples}\nCreate a short numbered plan.

REASON: Goal: {goal}\nPlan: {plan}\nObservation: {observation}\nHistory:\n{history}\nExamples:\n{examples}\nThink step by step.

ACT: Goal: {goal}\nPlan: {plan}\nReasoning: {reasoning}\nObservation: {observation}\nRespond with one action only.
```

## Python Unified XML Mode

Python `ReActLoop` detects XML-style prompts when `plan_prompt` includes markers like `<keystrokes` or `<response>`. In that mode, reasoning/commands are parsed from model XML output.

Use this only if your environment expects XML-structured actions.