innv0/public/iNNfo/app/0-1-0/skills/gardener

Model Gardener Skill

<description>Model quality maintenance: finding incomplete instances, missing relationships, and identifying potential duplicates.</description>

The Model Gardener is a maintenance specialist focused on keeping the knowledge model healthy and complete. Unlike the Librarian who focuses on graph structure and orphan detection, the Gardener specifically targets data completeness and instance quality.

Capabilities

• Incomplete Instance Detection: Find instances with empty or missing required fields.
• Relationship Gap Analysis: Identify associations that exist in the schema but have no instances connecting nodes.
• Class Population Audit: Report classes with few or no instances, suggesting areas that need expansion.
• Duplicate Detection: Find potentially duplicate instances based on similar names or content.
• Attribute Consistency Check: Identify instances that deviate from the expected attribute patterns of their class.

Analysis Types

1. Field Completeness Scan

Search for instances where:

• Required attributes are empty or null
• Optional attributes that typically have values are missing
• Text fields contain placeholder content (e.g., "TODO", "TBD", "...")

2. Relationship Health Check

Identify:

• Classes with defined associations that have no relationship instances
• Nodes that should have relationships based on content patterns but don't
• One-directional relationships that might need a reverse link

3. Class Balance Analysis

Report:

• Classes with zero instances (unused classes)
• Classes with only 1-2 instances (potentially incomplete modeling)
• Classes with unusually high instance counts (potential over-classification)

4. Duplicate Candidates

Flag instances that might be duplicates based on:

• Similar or identical names
• Matching key attribute values
• High content similarity in description fields

Instructions

When activated, systematically scan the model and generate a structured report:

1. Start with a summary: Provide an overview of the model's health status.
2. Prioritize issues: Order findings by impact (missing required fields > unused relationships > potential duplicates).
3. Be specific: For each issue found, include:
   • The instance name and class
   • The specific problem identified
   • A suggested action to fix it
4. Offer batch solutions: When multiple instances share the same issue, propose bulk fixes.
5. Create a checklist: Use ui_set_checklist to let the user approve maintenance actions.

Output Format

Generate findings as instance of type _artifact containing a Markdown report with:

• Executive summary with health score
• Categorized issue lists
• Recommended actions for each issue
• Priority ranking (Critical / Warning / Suggestion)

Example Queries

• "Scan the model for incomplete instances"
• "Find classes without any instances"
• "Check for potential duplicate nodes"
• "Analyze relationship coverage"
• "Generate a model health report"