Skafold Documentation

Working With AI

Import an AI-planned project

Create a Skafold project from a validated AI-generated seed, then use AI Context for future updates.

Skafold Project Importer turns confirmed planning into a skafold-project-seed file. A seed creates a new project; it is not used to update an existing one.

Generate the seed

Use the installed Skafold Project Importer in ChatGPT or Codex, or the standalone skill in another compatible AI tool.

Ask the AI to develop and confirm the plan before creating JSON:

Use Skafold Project Importer to develop this project plan. Keep assumptions and open questions separate. Do not generate the import file until I confirm the plan is ready.

When ready:

Create one validated Skafold project seed from the confirmed plan.

The skill reads its bundled schemas and guidance before generation.

Create the project

  1. Open the Skafold dashboard.
  2. Choose Import AI Plan.
  3. Select the generated .json file.
  4. Review validation errors and warnings.
  5. Confirm the import.

Skafold creates native editable data for Project, Structure, User Journeys, Sections, Page Hierarchy, Technical Architecture, Design Guidance, and Pre-flight.

Review the initial result

An AI-generated seed is a starting plan. Confirm:

  • Page names, paths, and direct parents.
  • Reusable Section definitions and groupings.
  • Page-specific placements, order, and instance names.
  • Assumptions and unresolved questions.
  • Technical or design decisions that need stakeholder approval.

Use Pre-flight to review imported delivery, QA, and launch requirements.

Continue with AI Context

After the project exists, stop exchanging whole-project seed files. Create one AI Context link from Project Tools → AI Context.

For every later AI revision:

  1. Publish the latest AI Context when Skafold reports updates are available.
  2. Give the same URL to the AI.
  3. Request the required frame or frame combination.
  4. Import the returned skafold-frame-update.

See Update an existing project with AI.

Shared Sections on import

Project seeds define each reusable Section once in globalSections. Page placements reference definitions with globalSectionId.

  • instanceName provides a page-specific label without renaming the definition.
  • Definition and placement status colours remain independent.
  • Divider rows use kind: "separator"; grouped definitions reference the divider through groupId.
  • Stable IDs allow future focused updates to preserve relationships.

Validation warnings

A warning does not always block import. For example, a page may intentionally begin without placements. Errors must be corrected before import and normally identify an unsupported field, duplicate ID, missing reference, invalid hierarchy, or malformed value.