Convert user story/task into executable PRP with deep codebase analysis
Transforms user stories into executable implementation plans by analyzing your codebase for patterns, dependencies, and integration points. Use this when starting development on new features to ensure your PRP includes all necessary context and validation steps.
/plugin marketplace add rpiplewar/shipfaster/plugin install prp@rapid-shippingTransform a user story or task into a tactical implementation PRP through systematic codebase analysis and task decomposition.
We do not write any code in this step, the goal is to create a detailed context engineered implementation plan for the implementation agent.
Key Principle: We must first gather the context about the story/task before proceeding with the analysis.
When we understand the story/task, we can proceed with the codebase analysis. We systematically dig deep into the codebase to gather intelligence and identify patterns and implementation points. We then use this information to create a PRP that can be executed by a coding agent.
The contents of the created PRP should encapsulate all the information the agent needs to complete the story/task in one pass.
Remember that subagents will only receive their details from you, the user has no way of interacting with the subagents. so you need to share all the relevant context to the subagent in the subagent prompt and in the TODO that is shared with the particular agent.
Create detailed todos and spawn parallel subagents to analyze (Use specialized subagents when apropriate):
Analyze the story to determine:
Get a deep understanding about the story/task before proceeding so that you can effectively guide the rest of the process.
1. Project Structure Analysis
2. Pattern Recognition
3. Dependency Analysis
4. Testing Patterns
5. Integration Points
Really think hard about everything you just learned during the research phases.
Transform analysis into concrete tasks:
Read and understand the template @PRPs/templates/prp_story_task.md
Task Rules:
Task Action Types:
We use the concept of information dense keywords to describe the action to be taken, below is a guidance. But you can use your own words to describe the action to be taken as long as you follow this same principle.
Examples:
For each task, design validation that:
Save as: PRPs/working-memory/{feature-name}/story_{kebab-case-summary}.md
Implementation Ready: Another developer could execute these tasks without additional context Validation Complete: Every task has atleast one working validation command Pattern Consistent: Tasks follow existing codebase conventions