seasoned-skills

Seasoned's skills to orchestrate the full product cycle with AI agents.

This package carries one complete way of working, run inside Claude Code, Anthropic's coding agent: how an idea becomes a shaped project (one written down and thought through before any build starts), how AI agents build it end to end, with everything that needs taste running on Fable, the strongest Claude model, and how the lessons learned along the way are kept. A project adopts it by installing the package and answering an interview once. From then on, everything the workflow needs inside the project is generated from the installed version: the standing instructions, the skills, the small supporting tools. The project commits only its configuration and its own content, and staying current means upgrading a version.

We built it because we were maintaining the same way of working in three repositories, by hand. Every lesson was learned once but taught to only one project, and the copies quietly drifted apart. Now a lesson written once reaches every project on its next upgrade, a new project costs an install instead of an afternoon of copying and pruning, and differences between projects are configuration, never forks.

Where this is going

We are building toward one thing: AI agents doing all the work of product development. Pulling requirements out of meetings, issues, user feedback, and the running product's behavior. Shaping, building, releasing, and watching what ships. Today we still bring the ideas and rule on the quality. Real autonomy means agents decide what to shape and shape it themselves, then decide what ships, to which audience, and how fast the rollout goes. Feedback from the outside world — the running product observed, users talking to the AI itself — starts the next shaping session on its own. Two doors stay open to us for good: anyone can start a shaping session, and every shaping document waits for human review before the build starts. But coming up with the work stops being something only we do.

Find your way in

• Understanding the way of working. The way of working tells the whole story, from the moment an idea arrives to the moment it ships. Running a session is the manual for the person at the keyboard while agents build.
• Adopting it. Adopting the workflow covers what your codebase needs in place first, the install itself, and how to stay current afterwards.
• Looking something up. The reference answers in under a minute: the commands, the configuration file, what a project receives, and the files and settings the package manages.

This repository

This repository is consumer number zero: the workflow it ships is installed here from the package itself, so the repo works under the very rules it distributes. It holds itself to the same adopter's bar the docs state, in the form that applies to a command-line package: its own CI runs both test tiers as required checks.

License

MIT.