mcp-engine-ai-readiness

name: mcp-engine-ai-readiness description: Assess Power BI semantic models for Copilot, Fabric data agent, and natural-language Q&A readiness. Use when reviewing whether a model has clear business terminology, unambiguous metrics, usable date defaults, focused field exposure, descriptions, AI instructions, AI data schema recommendations, verified-answer candidates, or natural-language validation tests.

PBI AI Readiness

Use this skill to turn an existing Power BI semantic model into a Copilot-ready assessment and artifact pack. This is an authoring and review workflow, not a runtime MCP tool.

Start Here

1. Confirm the user wants a readiness assessment, artifact drafts, or both.
2. Prefer metadata-level inspection before querying data values.
3. Use existing SemanticOps MCP tools when available: list_model, manage_dependencies, run_query, manage_tests, and manage_model_properties.
4. Keep unsupported Prep data for AI actions as drafts for Power BI Desktop, Power BI service, PBIP, Git, or manual review.
5. Separate recommendations into:
   • can apply through MCP/model metadata now
   • draft/export for Prep data for AI UI or PBIP/Git workflow
   • validate manually in Copilot

Workflow

• Read copilot-readiness-workflow for the assessment sequence, tool usage, privacy guardrails, and output order.
• Read readiness-scorecard when producing severity, score, business impact, and remediation priority.
• Read ai-artifact-templates when drafting AI instructions, AI data schema recommendations, verified answers, or manage_tests candidates.
• Read domain-examples when the model is sales, finance, support, or operational and the user wants concrete starting examples.

Guardrails

• Do not claim SemanticOps MCP can directly configure all Power BI Prep data for AI settings over live TOM/XMLA.
• Treat AI instructions, AI data schemas, and verified answers as draft artifacts unless the user provides an explicit supported PBIP/Git path or asks for manual-application guidance.
• Do not expose sensitive values from data previews. Prefer names, descriptions, expressions, relationships, dependencies, and aggregate-only validation queries.
• Respect SemanticOps MCP mode, policy, confirmation, license, and audit gates for any suggested or requested model change.
• Make nondeterminism explicit: readiness work can improve Copilot behavior, but it cannot guarantee identical answers for every prompt.

Output Standard

Return a compact readiness pack unless the user asks for raw details:

1. Executive summary with readiness level.
2. Scorecard grouped by critical, high, medium, and low findings.
3. Recommended MCP-applicable model metadata fixes.
4. Draft AI instructions.
5. Draft AI data schema recommendation.
6. Verified-answer backlog.
7. Optional natural-language test suggestions.
8. Manual validation checklist for Power BI Desktop or service.