One brief. Better build. Proof included.
Brief2Ship is a lean build standard for AI-assisted product work. It turns a strong one-shot brief into a shipped, verified artifact without making the user babysit a bloated process.
Brief2Ship is a workflow standard, not a giant framework.
It helps an operator or agent do five things consistently:
- Get a strong one-shot prompt
- Choose a base before building greenfield
- Reject brittle unreadable agent code
- Format reports and documents for decision-ready reading
- Turn the brief into a verified artifact
The external UX stays simple:
- One-shot prompt
- Up to 5 follow-up questions only if needed
- Execute
- Return proof
The user should not have to learn the internals.
The public name is Brief2Ship. The operating model underneath stays intentionally simple: one strong brief, three internal tiers, and a final proof gate.
Search for the best existing repo, starter, tool, or template before building.
Default companion: repo-first-starter
Use droogans/unmaintainable-code in reverse.
Reject:
- vague naming
- hidden logic
- duplicated business logic
- unjustified abstraction
- weak observability
- dependency bloat
- chat-context-only code
Do not ship generic AI-slop UI.
Default references:
shadcn-ui/uitailwindlabs/headlessuimotiondivision/motiondarkroomengineering/leniswhen justifiedmagicuidesign/magicuiselectively only
Most starter kits solve only one part of the build problem:
- scaffolding
- design system
- SaaS boilerplate
- component library
Brief2Ship is different.
It is a build operating standard for AI-assisted delivery:
- choose better starting points
- preserve judgment and maintainability
- add tasteful finish without adding ceremony
The old names described the internal mechanism or the final proof. Brief2Ship sells the full transformation:
- Brief — the user gives one clear starting prompt, not a pile of process docs
- 2 — direct path from intent to execution, with at most five targeted follow-ups
- Ship — the result is a usable artifact with receipts, not a planning loop
- Memorable + searchable — short enough for prompts, docs, repo names, and package names
Use this line when explaining it publicly:
Brief2Ship is the lightweight operating standard for AI-assisted builds: start from one strong brief, choose the best base, keep the code maintainable, polish the UX, and prove it works.
skills/brief2ship/SKILL.md— Hermes/OpenClaw skill packagedocs/how-it-works.md— full breakdown of the workflowdocs/lanes.md— the 4 default lanesdocs/examples.md— example prompts and execution patternstemplates/— kickoff, report, and receipt templatesscripts/install-skill.sh— install the skill into a local Hermes profile
To avoid ambiguity without overcomplicating the standard, this repo ships with four default lanes:
- App
- Dashboard / Internal tool
- Landing page
- Report / Document
See docs/lanes.md and docs/report-document-lane.md.
Every build should end with a small artifact pack:
- chosen base and why
- changed files
- commands/tests run
- verification/proof
- known compromises
Use templates/build-receipt-template.md.
This project intentionally does not try to be:
- a full app framework
- a replacement for product judgment
- a giant scoring engine
- a mandatory design system
- a rigid PM workflow
The goal is to improve build quality without harming judgment or accuracy.
| Product type | What it does well | What this adds |
|---|---|---|
| Starter kits | Scaffold fast | Forces repo-first choice instead of default greenfield |
| SaaS boilerplates | Ship common product features | Adds maintainability gate + anti-slop finish rules |
| Component libraries | Better UI primitives | Adds upstream selection + delivery discipline |
| Agent skills/runbooks | Internal process | Adds a public-facing packaging layer and examples |
Read:
bash scripts/install-skill.shThen load:
brief2ship
Backward-compatible skill names during migration: shipproof, three-tier-build-toolkit.
See CONTRIBUTING.md.
Initial framework direction and product requirements: Mike
Packaging, workflow codification, and repo drafting: Hermes Agent (generalist1)
Future contributors should keep the project lean. If a feature makes the workflow harder to apply consistently, it probably does not belong here.
MIT for the material in this repo. Upstream referenced projects keep their own licenses.