A brand-free, portable operating model for running software projects with Claude across its two surfaces: a coordination/reasoning surface (Claude.ai) and an agentic build surface (Claude Code). This is the distilled, reusable core of a methodology refined in production and published so others running similar projects can adopt it instead of starting from a blank page.
methodology/Portable-Operating-Model.md— the backbone. Separates why a practice exists (principle) from how one project did it (instantiation), so you adapt rather than cargo-cult. Start here.methodology/PI-core.md— Project Instructions (generic core): the always-on disciplines a coordinator and build team carry in-thread — the Story/Brief split, hard-stop rules, verification discipline, defense-in-depth, branching/PR workflow, orchestration-primitive selection.methodology/CLAUDE.md— the in-repo distillation of the PI: the lean, always-loaded operating contract for a Claude Code build session.
It names no project, person, or product by design. You add your own project layer (domain, stack, invariants) on top.
Solo operators and small teams running Claude-based software projects who want a tested methodology, especially anyone coordinating work across Claude.ai and Claude Code.
See ADOPTING.md. In short: read the Portable Operating Model, drop PI-core.md and
CLAUDE.md into your project, then layer your project-specific facts on top.
This is a marked snapshot — methodology as of the version stamped in CHANGELOG.md. It is
re-published on call when the underlying method moves, not continuously mirrored.
Released under CC BY 4.0: use, adapt, and build on it (even commercially); just credit
the source. Attribution: <your-name-or-handle> (replace before publishing).