append-prd

Convert requirements into actionable dev work items appended to a project's plans/prd.json.

Inputs

• •Requirements: PRD doc, raw ideas, bullet points, plan -- any form
• •Repo: the project repo to ground tasks in

Process

1. •Read/parse the provided requirements
2. •Explore the repo -- understand architecture, key files, frameworks, existing code patterns
3. •Break requirements into smallest actionable dev tasks
4. •Append tasks to {project}/plans/prd.json

Task schema

Each task in prd.json follows this structure:

{
  "category": "functional | ui | api | data | auth | performance | ...",
  "description": "Clear outcome-oriented description of what to build/change. Enough context for a dev to pick up independently.",
  "steps": [
    "Actionable step a dev follows to complete this task"
  ],
  "passes": false
}

prd.json is a top-level JSON array of these objects.

Rules

• •Explore first: always read the repo before generating tasks. Reference real files, components, routes, functions.
• •Append, don't overwrite: read existing prd.json first, parse it, then append new tasks to the array.
• •Create if missing: if plans/prd.json or plans/ dir doesn't exist, create them. Initialize with [] before appending.
• •All tasks start unverified: every task gets "passes": false.
• •Small tasks: each task should be the smallest unit of dev work someone can pick up and complete. Prefer many small tasks over few large ones.
• •Clear steps: each step is an actionable instruction a dev can follow to execute and complete the task. Reference actual code paths, components, endpoints.
• •Appropriate categories: use functional, ui, api, data, auth, performance, or other fitting categories. Match the nature of the task.
• •No duplicates: if appending to an existing prd.json, check for duplicate/overlapping tasks and skip them.