Pdd Prompts

# Generate a Feature Prompt

This is the Claude adapter for the shared `Prompts` workflow in `core/workflows/prompts.md`. Keep shared workflow behavior aligned there; this file exists to preserve Claude-specific command wording and `$ARGUMENTS` handling.

**User input**: $ARGUMENTS

## Purpose

Create focused, versionable prompt artifacts that ask the AI to do one job well.

Prompts should be durable project artifacts, not one-off chat messages. Each prompt should have a single job, a clear file path, and a reviewable output expectation.

## Use When

- The team is ready to implement a planned change.
- Context is already in place or good enough to proceed.
- A feature is small enough to prompt directly or already decomposed into clear phases.

## Inputs

- feature description
- current context files
- plan or research notes if they exist

## Step 1: Load Context

Read the relevant project context first. If `pdd/context/project.md` exists, use it. Pull in conventions and decisions if they affect the requested feature.

Load the matching project-type reference file for prompt patterns, common feature concerns, and type-specific expectations.

If context is missing, continue but note that a later `/project:pdd-context` pass would improve prompt quality.

## Step 2: Decompose If Needed

Watch for work that should be split before writing prompts:

- multiple unrelated outputs
- backend and frontend bundled together
- implementation plus tests plus refactor all in one ask

If the task is too broad, split it into smaller prompts before writing anything.

## Step 3: Gather Task Details

Ask for the details needed to make the prompt precise:

- what the feature does
- inputs and expected outputs
- constraints and edge cases
- surrounding code or modules it must fit into

## Step 4: Write The Prompt

Before writing from scratch, check whether an existing template in `pdd/prompts/templates/` already fits.

Use a structure like: