@xenonbyte/da-vinci
v0.1.1
Published
Requirement-to-design-to-code workflow skill for Codex, Claude, and Gemini
Downloads
87
Readme
Da Vinci
Da Vinci is a software delivery workflow that turns product requirements into structured specs, Pencil designs, page-to-design bindings, implementation tasks, and final code.
The workflow is built around one fixed contract:
- requirements decide behavior
- Pencil decides presentation
- code follows both
What Da Vinci is for
Use Da Vinci when software delivery needs all three layers to stay aligned:
- requirement breakdown
- Pencil-backed UI design
- implementation from requirements plus design data
This workflow is intended for:
- 0-to-1 product delivery
- brainstorm-to-product delivery
- redesign from existing code
- scoped feature delivery on an existing product
Supported workflow modes
Da Vinci V2 supports four modes.
greenfield-spec
Use when the project is new and the requirements are already clear enough to write specs.
greenfield-brainstorm
Use when the project is new but the inputs come from multiple rounds of rough product ideation and need synthesis first.
redesign-from-code
Use when an existing project already has routes, pages, and UI, and the goal is a broad redesign or UI replacement.
feature-change
Use when an existing product needs a scoped requirement change, page addition, or page update.
Core workflow
Da Vinci runs in this order:
- select the active mode
- build the correct source artifacts
- collect design inputs and register design sources
- define or discover the project page map
- create or update Pencil designs
- bind implementation pages to Pencil pages
- read Pencil design data through MCP
- generate implementation tasks
- implement code from requirements plus Pencil data
- verify requirement drift and design drift
Default artifacts
Depending on the mode, the workflow may use these artifacts:
brainstorm.mdproject-inventory.mddesign-brief.mddesign-registry.mdpage-map.mdproposal.mdspecs/<capability>/spec.mddesign.mdpencil-design.mdpencil-bindings.mdtasks.mdverification.md
Artifact intent
brainstorm.md: raw ideas before specs are stableproject-inventory.md: current routes, pages, and UI patterns from an existing codebasedesign-brief.md: form factor, style, density, brand constraints, and layout preferencesdesign-registry.md: project-level.pensource inventorypage-map.md: canonical page list and page responsibilitiesproposal.md: change scope and outcomesspec.md: behavior, states, edge cases, and acceptance rulesdesign.md: page structure, interactions, and component strategypencil-design.md:.penfile path, screens, screenshots, and implementation notespencil-bindings.md: implementation page -> Pencil page bindingstasks.md: implementation order and task groupsverification.md: requirement coverage, design coverage, and drift findings
Checkpoints
Da Vinci uses these checkpoints:
discovery checkpointspec checkpointdesign checkpointmapping checkpointtask checkpointexecution checkpoint
Checkpoint outcomes:
PASSWARNBLOCK
Execution Policy
Da Vinci is autonomous-by-default.
That means:
- it should continue the workflow without pausing for user confirmation at each stage
- checkpoints are internal quality gates, not approval gates
- it should only stop when a real blocker exists
Default behavior:
PASS-> continueWARN-> record and continueBLOCK-> stop only when the blocker cannot be resolved from existing artifacts or local context
Typical stop conditions:
- the user explicitly interrupts
- a missing input would materially change the design or implementation result
- an external permission or authentication step is required
- source artifacts conflict badly enough that continuing would guess the truth
- a destructive baseline-changing action is about to happen
Design source policy
Da Vinci treats Pencil as a structured design source, not as a screenshot source.
Rules:
design-registry.mdis the project-level inventory of.pensourcespage-map.mdis the source of truth for implementation pagespencil-bindings.mdis the source of truth for implementation page -> Pencil page mapping
When a relevant mapping already exists:
- iterate on the mapped Pencil source
When mappings do not exist but local code exists:
- reconstruct the baseline from the local project
- rebuild inventory, page map, and design registry
When neither mappings nor usable design sources exist:
- create a new Pencil baseline from the current local source of truth
Design input collection
Before generating Pencil designs for a greenfield project, Da Vinci should collect or infer:
- product form factor
- visual tone
- density
- brand direction
- responsive priority
Preferred inference order:
- existing artifacts
- project-local signals
- explicit user answers
- short clarification questions
Installation
Install the package:
npm install -g @xenonbyte/da-vinciInstall platform assets:
da-vinci install --platform codex,claude,geminiUseful commands:
da-vinci status
da-vinci validate-assets
da-vinci uninstall --platform codex,claude,geminiInstallation targets:
- Codex prompts:
~/.codex/prompts/ - Codex skill:
~/.codex/skills/da-vinci/ - Claude commands:
~/.claude/commands/ - Gemini commands:
~/.gemini/commands/
Platform usage
Codex
Preferred:
$da-vinci <request>Claude
Preferred:
/da-vinci <request>Gemini
Preferred:
/da-vinci <request>Repo-local forward test
A complete repo-local forward test is available at:
examples/greenfield-spec-markupflow/
That example runs a greenfield-spec scenario from:
- design brief
- proposal and spec
- page map
- Pencil source registration
- Pencil page bindings
- implementation tasks
- static HTML delivery
- verification
Example requests
Greenfield spec
$da-vinci use greenfield-spec to break down a new desktop software project for annotated design deliveryBrainstorm synthesis
$da-vinci use greenfield-brainstorm to turn several product ideas into a page map, Pencil design plan, and implementation tasksRedesign from code
$da-vinci use redesign-from-code to inventory the existing app, identify current pages, and generate a new Pencil-backed UI planFeature change
$da-vinci use feature-change to add a billing page, generate the relevant Pencil design delta, and create implementation tasksWorkflow examples
See:
docs/workflow-examples.md
Repository layout
da-vinci/
├── SKILL.md
├── agents/
│ └── openai.yaml
├── commands/
│ ├── claude/
│ ├── codex/
│ └── gemini/
├── references/
└── openspec/Current status
The repository currently contains:
- the Da Vinci skill
- three-platform command adapters
- workflow references
- OpenSpec planning artifacts for the skill itself
- workflow examples
Next natural work for the repository:
- add install/distribution helpers
- forward-test the workflow on a real project
