Generate E2E tests from a described user flow

The problem

AI coding agents lose context between sessions, forcing you to re-explain your tech stack, folder structure, and coding conventions every time you open a new chat. Manually briefing an agent on a large repo is tedious and prone to missing critical architectural decisions.

What it does

• Scans your project root to identify the primary language, framework, and package manager.
• Extracts coding conventions directly from linter configs like Prettier, ESLint, and Rubocop.
• Maps the top-level directory structure while ignoring noise like node_modules and build artifacts.
• Surfaces recent git history and ADRs to give the agent historical context on recent changes and design choices.
• Generates a portable PROJECT_MEMORY.md file that serves as a universal brief for any LLM.

Frameworks & tools

Supports JavaScript/TypeScript (Node.js, Next.js), Python (FastAPI, Django), Rust (Cargo), Go, Ruby (Rails), PHP (Composer), and Java (Maven).

Why this beats prompting it yourself

This skill systematically parses configuration files and directory trees that you would likely skip or summarize incorrectly. It ensures the agent receives a standardized, structured brief that fits within context windows without leaking sensitive env secrets.

Use cases

• Onboarding an AI agent to a legacy codebase with non-standard conventions.
• Switching between different AI IDEs or chat interfaces while maintaining consistent context.
• Setting up a new project and establishing a "source of truth" for AI coding assistants.
• Updating project context for an agent after a heavy refactor or dependency migration.

Known limitations

Caps file reads at 4KB per file and directory depth at 3 levels to remain performant and context-efficient. It cannot read encrypted files or your actual .env secrets.