A reusable engineering setup for new projects - a professional process from commit one, with no lock-in.
DevBlueprint is a reusable engineering setup for new projects, extracted from a real production codebase and made stack-agnostic. One command scaffolds the git workflow, quality gate, code conventions and AI-assistant guardrails into a target directory. Documentation-first: the output is plain files you own and edit - no runtime, no lock-in. You can delete DevBlueprint afterwards, and nothing breaks.
Every new project starts without a process: the git workflow, quality gate, conventions and AI guardrails get reinvented every time or skipped entirely - especially solo. Running several AI sessions in parallel adds a second problem: the chats keep switching each other's branch. Ready-made frameworks and template repos only solve it at the price of permanent lock-in.
Everything - the idea, extracting the process from a real production codebase, the stack-agnostic core docs, the Bash CLI, the nine stack variants, the bats test setup and the agent integration (CLAUDE.md, --json, setup skill).
DevBlueprint separates a stack-agnostic core from thin stack overlays. core/ holds the source of truth - the git workflow, engineering standards, conventions, quality-and-testing and the CLAUDE.md templates. Nine variants add only the stack-specific parts: the concrete gate commands, a CI workflow, .gitignore and the worktree configuration. A Bash CLI scaffolds, checks and updates the setup.
core/ is the stack-agnostic source of truth and is copied into every project; variants/ only overlay the stack-specific parts. The output is plain files in the target project - there is no runtime and no dependency on DevBlueprint itself.
A shared setup as a framework or template repo binds every project to it permanently - updates become a dependency, deviations a fight against the tool.
Documentation-first: the output is plain files the project owns and edits freely. There is no runtime; after scaffolding, DevBlueprint can be deleted without anything breaking.
Several assistant chats on one repo switch each other's branch away - work lands on the wrong state or is lost.
One worktree per task via a single wt script: each task gets its own directory with its own branch (off develop), the main clone stays on master. Sessions no longer collide.
The process should be the same everywhere, but the concrete quality-gate commands, CI and ignore rules differ fundamentally by stack.
core/ holds the agnostic core, variants/ overlay only the stack-specific parts. A new variant is created by copying a folder and adjusting the gate, wt.conf and CI - it then appears automatically in devblueprint list.
A text CLI is built for humans; an agent must parse the state reliably and confirm before every write.
list, doctor and version emit --json (doctor still exits non-zero on errors), an intake.yml plus plan (== init --dry-run) shows exactly the planned output, and a /devblueprint-setup skill interviews, writes the intake file and scaffolds only after confirmation.
Started as an agent project kit with a CLI scaffold, first docs and CI. Five commits lay the base before the concept is reworked.
Rebuilt into a documentation-first engineering kit in one concentrated workday: the core docs, the worktree-based workflow, nine stack variants, the CLI commands (detect, --json, intake + plan, update), the Bats test suite and the agent setup skill. 112 of the 117 commits fall on this day - including dogfooding the setup on its own repo.