Deep Work Plan is an open-source, agent-agnostic methodology that turns any repository into a structured environment with context, guardrails, and a durable plan so any coding agent can execute with precision on long-horizon work. It addresses agent drift, where context fills up, decisions are forgotten, and multi-hour tasks get abandoned, by writing the spec directly into the repo with atomic tasks, acceptance criteria, validation gates, and resumable state. Long runs survive context resets, so any agent can pick up where the last one left off. A DWP-verify step produces an objective pass/fail report against the specification, so AI-first work is verified rather than asserted. It works with Claude Code, Codex, Cursor, or any other agent, has no lock-in, and is released under the MIT license.