Load a hierarchy and generate navigation diagrams

Goal

Load a hierarchical reference model, such as a capability map or technology taxonomy, into the repository and build click-through drill-down diagrams over it. For an architect who owns a reference model. You end with one landing diagram and one child diagram per parent element, with each parent flagged as having a child diagram, verified by an audit and by a render.

Where it runs

This scenario needs the MCP server. Computer use covers its UI steps.

Before you start

• The start-of-session check done.
• A source for the hierarchy: a spreadsheet, a catalog export, or elements already in the model. The levels and the parent of each element must be recoverable.
• A hierarchy that meets the skill's threshold: at least two levels deep, more than about twelve elements, and at least three elements with children. Below that, one diagram is better. The hierarchy should be stable, because the set must be rebuilt when it changes.
• A target package, and a baseline of it.
• Computer use, so that the AI can run the double-click test in EA's window, or a person at EA. It is the only check that exercises navigation.

A capability model, a technology reference model or an organization structure is data until somebody makes it browsable. A single tree diagram stops being readable past about thirty elements. A navigation diagram attaches a diagram to an element as its child and flags the element. EA draws a small marker in the element's corner, and double-clicking opens the child diagram. Chained across levels, a landing diagram of the top level opens the level below it, down to the leaves. ea-navigation-diagrams builds the set. After a load that creates a qualifying hierarchy, or an analysis that reveals one, the skill offers to build it. It is not a hyperlink, a diagram frame or a navigation cell. Those are different EA features.

Steps

1. AI with computer use; otherwise you If EA is not running or the model is not open, start EA and open it. AI Power Tools ea-start-here runs the start-of-session check.
2. You Confirm the target model.
3. AI ea-modeling reads the source and states which hierarchy it will treat as the source of truth: element parentage, containment connectors, or a level tag. A model can hold the hierarchy in more than one place, and they can disagree.
4. You Choose the source of truth, and review any disagreement the AI reports. EA's Browser and nesting render from element parentage, so prefer it.
5. AI Power Tools ea-modeling loads the hierarchy: the packages and the elements in bulk, with each element's parent recorded.
6. AI Power Tools ea-navigation-diagrams checks the threshold and what is already navigable (find_composite_diagram_mismatches). It reports a child diagram without the flag (it looks navigable and does not drill down) and a flag without a diagram (a marker that opens nothing). If every qualifying parent is already navigable, the skill says nothing.
7. You Choose what the skill does: build the set, complete an existing set, fix only the flags, or do nothing. Declining is remembered for the rest of the session. Fixing only the flags is the cheapest choice.
8. AI Power Tools ea-navigation-diagrams creates one child diagram per parent and places a nested layout. Each child diagram shows the parent drawn as a large container with its immediate children inside it. There are no connectors: the hierarchy is shown by nesting, and connectors would rebuild the tree diagram this replaces. The skill computes the geometry: a 130 by 80 child box, 40-unit gaps, up to four columns, and the container sized around its children. The landing diagram is the same grid with no container. The container sits behind its children.
9. AI Power Tools ea-navigation-diagrams links each parent to its child diagram (set_composite_diagram), which writes both halves of the link and reads them back. A response that reports success without consistency means the write did not fully land.
10. AI Power Tools ea-navigation-diagrams audits and counts. Expect: diagrams equal one plus the number of parents, objects equal the roots plus, for each parent, one plus its children, no unstyled objects, no stray links, and flagged elements equal parents.
11. AI with computer use; otherwise you Reload the package in EA's Browser (right-click the package, Contents, Reload Current Package), because writes to the repository do not reach an EA session that has already cached the package. Then look at the landing diagram, a mid-level diagram and a leaf-level diagram. Counting rows is not verification. The deliverable is a picture, and you judge it.
12. AI with computer use; otherwise you Double-click a landing box and a mid-level box in EA, and confirm that the child diagram opens.
13. AI ea-navigation-diagrams reports the set and names the entry-point diagram.

Prompts to use

Opening

Follow-ups

The double-click test, with computer use.

Recommended models

Ratings are guidance based on the kind of reasoning the skill needs, not measured benchmarks.

Checking the result

• Accurate. The audit over the scope returns no mismatches, and every link write reports consistent. The three renders show the marker on every box that should drill down, children inside the container, and no clipped names. A person double-clicks a landing box and a mid-level box, and the child diagram opens.
• Complete. The counts match: diagrams equal one plus the parents, flagged elements equal the parents. Leaves are unflagged and have no diagram. The report names the entry-point diagram.
• Nothing unintended changed. Comparing the model with its baseline shows the new package content and the new diagrams. Existing diagrams are not deleted or overwritten. Anything superseded is moved to a holding package inside the same scope, and you are told what moved.

Variations

• Fix only. Run only the threshold check, the choice to fix only the flags, the linking and the audit, on a set someone hand-built. A hand-built set often has both halves wrong in different elements.
• Complete an existing set. Read an existing navigation diagram's type and style and copy them, so new diagrams match.
• Other hierarchies. Technology reference models, product taxonomies, organization structures, requirement and risk breakdowns, data domain models and value stream hierarchies fit the same pattern.
• Rebuild after a change. A new parent needs a diagram and a flag. A parent that lost all its children should have its diagram removed and its flag cleared. A renamed element needs its child diagram renamed. The cheapest correct approach is to regenerate the set for the affected package.

Limitations

• Drill-down may need checking in EA. If the marker is drawn but double-click opens Properties, the navigation is not working on your build. Report it (Report a problem to support) and take a baseline before trying any workaround.
• The server's own verification does not test navigation. Counting, auditing the two halves and rendering all pass with the marker drawn. Only the double-click test, done in EA's window by the AI with computer use or by you, exercises the behavior.
• EA can still draw relationships on a child diagram. Telling AI Power Tools not to write links stops it writing link rows. EA draws a relationship between two placed elements whether or not a row exists. Placing with no layout guarantees no stored link.
• Duplicate sibling names are legal in EA and common in imported taxonomies. Their diagram names duplicate too. This is harmless, and the report says so.
• Long names need wider boxes. A 130 by 80 box holds about four short words. The AI measures the longest name before fixing the width.
• The Browser needs a reload. Writes to the repository do not reach an EA session that has the package cached.
• Navigation diagrams are derived artifacts. Rebuild the set when the hierarchy changes.