Are Skills the New Apps? (elite-ai-assisted-coding.dev)

An AI user converted their CardCraft flashcard web app into a reusable “skill” by feeding the entire codebase (concatenated with the files-to-prompt CLI) into Claude and asking it to repackage the app as a single downloadable skill. Claude parsed the source code and prompt templates, restructured the card-generation logic and Python Anki-exporter, and produced a working skill without any new code being written. A live test — creating flashcards from the Wikipedia article on the French Revolution — produced recall-focused cards and an import-ready Anki package, demonstrating end-to-end functionality. This demonstrates a shift toward “skills” as lightweight, portable app abstractions: instead of managing infrastructure and deployments, developers can let LLMs ingest implementation details and emit standardized skills that run in Claude web, Claude Code, or via integration (e.g., a Skillz MCP server) and be shared through marketplaces like the Skills Supermarket. Technically, it highlights LLMs’ ability to read, refactor, and encapsulate prompt-based pipelines and export code artifacts, enabling faster prototyping and tighter agent integrations. Caveats include dependence on model access and skill-format standards, but for AI/ML practitioners this approach suggests a new pattern for packaging capabilities rather than shipping traditional apps.