PAD is a method for working with coding agents — Claude Code, Cursor, Codex, Gemini, any of them — that keeps a project's context inside the repository, where it stays portable, reviewable, and yours.
The project's context belongs to the project, not to the tool. — the root principle
Agentic tools change every quarter; a serious project outlives all of them. PAD puts the context an agent needs into the repository itself, so the work survives the churn — and any agent can pick it up without reconstruction.
The rules, the architecture, the decisions — in tool-agnostic files under /docs/, not in a vendor's memory. Thin adapters (AGENTS.md, CLAUDE.md, .cursor/rules) point every tool there.
Every behavioral change moves code, tests and docs together — so the Git history becomes a readable architectural log.
Think, Plan, Execute, Verify. The human approves before code is written; verification is visible, not assumed.
Small atomic commits on branches mean any step can be undone with git revert in seconds. Mistakes stop being expensive.
Read the method, put it to work for free, or get the tooling that generates and enforces it.
The full account of PAD — its origin, principles, the cycle, adapters and Skills, and how it differs from Scrum, XP and DevOps. Where to begin if you want the why.
PAD v1.8 · First edition, revised · November 2026 · 152 pages
Put PAD to work the same afternoon. The four deliverables promised in the book, unlocked with the code on its Reader toolkit page.
Launching soon · for teams and serious builders
The tooling that turns the method into a machine: generate a project's whole /docs/ with any assistant, and enforce the coupled commit in CI.
Openb helps teams install the method on their own repositories, tools and compliance constraints — beyond what any kit can do on its own.