prd_skill

Overview

This skill creates agent-friendly Linear tickets by producing clear PRDs, parent issues, and focused sub-issues with verifiable success criteria. It is designed to prepare work for human or agentic implementation and to enforce proper scope, dependencies, and verification steps.

How this skill works

I translate a feature request or problem statement into a concise parent issue (epic) and a set of actionable sub-issues. Each ticket includes problem/context, must-have and nice-to-have requirements, explicit success criteria, verification commands, and in/out-of-scope boundaries. I also set dependency relationships so work executes in the correct order.

When to use it

• Planning a new feature that will be implemented by humans or AI agents
• Breaking down a large task into focused, verifiable sub-issues
• Preparing tickets for agentic coding where clear verification is required
• Converting vague requests into testable, time-boxed work items
• Coordinating multi-team work that needs explicit dependencies

Best practices

• Size tickets for 1–4 hours of focused work so each is completable in a single session
• Always include concrete verification commands (tests, lint, build) that confirm success
• Be explicit about boundaries: list what is in scope and what must be split into separate tickets
• Label agent-ready tickets and search for existing related work before creating duplicates
• Use --depends-on and --blocked-by to model real-world sequencing and blockers

Example use cases

• Create a parent issue describing a new authentication system with sub-issues for OAuth, session handling, and migration
• Break a monolithic feature into independent sub-tasks each with its own success tests and verification commands
• Prepare a feature handoff for an AI agent by adding clear success criteria and commands to run locally
• Identify and link existing related issues before adding a new epic to avoid duplication
• Mark migration tasks as blocked by database schema updates and set dependencies accordingly