Seven skills for Claude AI: a writing baseline, a PR/FAQ generator, a full use-case-to-tracker-issue pipeline, and architecture recommendations.
Click below to open a new claude.ai chat, pre-loaded with the whole planning flow: Working Backwards PR/FAQ first, then use case discovery, use case writing, test cases, and requirements, as the conversation gets there. Uses your own claude.ai account and session.
Get started in claude.ai →This asks Claude to fetch skill files from GitHub as needed, so it needs a claude.ai plan with web fetching enabled. If it doesn't work, copy this SKILL.md link into a new chat yourself and ask Claude to follow it.
This opens with whatever model your account defaults to, often Opus. None of these skills need it; switch to Sonnet 5 in the model picker once the chat opens for a faster, cheaper run.
Working Backwards PR/FAQ is the natural entry point: it clarifies what you're building and why before anything else. From there, the other skills pick up the thread: breaking the idea into use cases, writing each one out, generating test cases, and deriving trackable requirements you can export to Jira or GitHub Issues. Once there's enough to work with, Architect Review turns it into a structured build recommendation.
Watch a full example run → — a real intake played back turn by turn, or skip straight to what it produced.
Full details, install instructions, and downloads: github.com/ToddE/claude-skills