IDX CMA Report

IDX CMA Report

Use this skill to turn subject-property data and IDX comparables into a defensible CMA package with:

• Structured valuation calculations
• A written report for agent/client review
• An interactive handoff prompt for Google Gemini Canvas / Google AI Studio

Workflow

1. Gather Data Through IDX MCP/CLI

Use the IDX MCP/CLI skill already available in the environment to pull:

• Subject property details
• Candidate comparable listings (closed/pending/active based on user preference)

Ask the user which comps to include when the choice is ambiguous. Keep 3 to 8 comps unless the user requests otherwise.

Normalize data to JSON using the schema in references/cma-input-schema.md.

2. Build CMA Outputs

Run:

python3 scripts/build_cma.py \
  --subject subject.json \
  --comps comps.json \
  --output-dir cma-output

The script produces:

• cma-output/cma_report.md (summary report)
• cma-output/cma_data.json (calculation payload)
• cma-output/interactive_local.html (local interactive view)
• cma-output/gemini_canvas_prompt.md (prompt for Google tools)

3. Review and Explain Adjustments

Before final delivery:

• Show the comp set used
• Show estimated range and central estimate
• Explain assumptions and major adjustments in plain language
• Flag missing/low-quality fields that weaken confidence

Use references/valuation-guidelines.md for adjustment defaults and confidence guidance.

4. Publish Interactive Version in Gemini

Use cma-output/gemini_canvas_prompt.md as the base prompt. Then:

1. Open Google AI Studio or Gemini Canvas.
2. Paste the generated prompt and provide cma_data.json.
3. Ask for an interactive CMA web app with:
   • Comp table with sorting/filtering
   • Map-ready data fields (if lat/lng present)
   • Value-range visualization
   • Notes panel explaining adjustments
4. Request hosted/shareable output if available in the chosen Google tool.

See references/gemini-canvas-publish.md for a copy-ready checklist.

Safety Rules

• Treat outputs as broker/agent CMA support, not a licensed appraisal.
• Surface data gaps, outliers, or stale comps before presenting a valuation.
• Never invent listing attributes; mark missing values as unknown.
• Keep a clear boundary between factual listing data and model assumptions.

References

• references/cma-input-schema.md
• references/valuation-guidelines.md
• references/gemini-canvas-publish.md