Core Workflow

Project lifecycle

Move from a brief to a shared plan, review milestones, implementation, and ongoing updates.

Keep Skafold current as decisions change. The live project is where people and computer-use agents edit; AI Context is the reviewed snapshot you publish for trusted tools to read.

1. Start the plan

Create a blank project, choose a relevant preset, or plan with your preferred AI tool. For AI-assisted discovery, use Skafold Project Assistant to distinguish confirmed requirements, assumptions, and open questions.

When ready, ask for a validated project seed and choose Import AI Plan on the dashboard. A seed creates a new project with editable Documentation, Structure, and Pages.

2. Write the brief in Documentation

Capture the problem, audience, goals, scope, constraints, and unresolved questions. Add sections such as Stakeholders, Research, or Content requirements when useful.

Use references for briefs and source material, colour palettes and typography for brand decisions, and checklists for review and launch requirements. Section titles become the document's navigation; search helps the team find details later.

3. Map the system in Structure

Create Page nodes for the routes and screens. Connect direct parent and child pages, then map important user journeys and technical dependencies on the same canvas.

Use Title and Paragraph annotations to explain each diagram. Keep connections meaningful and leave enough space for the plan to grow.

4. Compose Pages

Plan each page's sections from top to bottom. Open Sections inside Pages to manage reusable definitions, their content fields, layout patterns, and build guidance.

Reuse a definition where its purpose and content structure fit. Give page placements their own names and status where needed. Shared definitions and page placements track status independently.

5. Review together

Use live presence to coordinate editing, check Saved before handing work over, and review the actual content in the project overview. Create a Share link for clients who need a view-only project.

Record review or launch requirements in Documentation checklists. Save named milestones at approval points in each relevant workspace, such as Brief approved or Ready for implementation.

6. Publish AI Context

Choose AI Context → Manage AI Context in the main header, then Create AI Context link. Creation publishes the first snapshot. The link belongs to the project and is shared by authorised editors.

Treat the URL as a secret: anyone holding it can read the published context. Give it only to trusted people and tools. It is separate from the visual client Share link.

After reviewing saved changes, use AI Context → Update AI Context when available. Tools see the new snapshot when they fetch the same URL again; editing alone does not publish it.

7. Keep the plan aligned with implementation

When implementation changes an agreed requirement, update the relevant workspace and republish AI Context. A computer-use agent can make focused edits through the signed-in editor, inspect the result, and verify saving. An agent without computer use can propose manual changes.

Review the affected workspace's History to see what changed. Agent edits use the signed-in account's normal history, so save a milestone before a significant revision if you need a clear checkpoint.

Existing-project bulk AI update files are no longer supported. If the plan needs a substantial rebuild, generate a seed for a separate project and retain the original and its history.

8. Handoff and archive

Complete delivery checklists, review the page and shared-section statuses, and save the milestones worth retaining. Publish the final agreed AI Context if implementation tools still need it.

Archive the project from the dashboard when it is no longer active. Archiving keeps the content accessible with existing permissions, but automatic history still expires under its normal policy.