Property-Report-Skill-Judges

Agents — Orchestrator + Advisors

This skill is one orchestrator plus five domain advisor subagents. Each advisor runs independently (by its own trigger phrases) or is invoked by the orchestrator at the right stage of the buyer journey. All of them read and write one shared per-property state spine (properties/<slug>/.property/journey.json) so numbers stay consistent.

• Full design: ../docs/AGENT-ARCHITECTURE-DESIGN.md
• User-facing walkthrough: ../docs/How-To-Use-Guide.html

Orchestrator (repo root)

Component	Location	Role
Property Investigator	../SKILL.md	Skill entry point. Owns routing, the buyer-journey checklist, the main investigation report, the contract & deadlines, shared-state writes for stage/contract, and cross-agent reconciliation.

The orchestrator’s SKILL.md must stay at the repository root — a SKILL.md at the folder root is what defines the skill and its entry point. The advisors below live one level down, each in its own folder.

Advisor subagents

Advisor	Folder	Leads stage	Independent triggers (examples)
Neighborhood Advisor	neighborhood-advisor/	Neighborhood & value	“comps”, “what’s this home worth”, “neighbors’ details”
Comparison Advisor	comparison-advisor/	Choosing among homes	“which should I buy”, “best buy”, “A vs B”
Loan Advisor	loan-advisor/	Financing	“compare loans”, “review my Loan Estimate”, “should I pay points”
Inspection Advisor	inspection-advisor/	Option period	“review my inspection”, “estimate repairs”, “should I walk away”
Insurance Advisor	insurance-advisor/	Insurance (pre-closing)	“compare home insurance”, “insurance call prep”

Shared contract

• Routing: match the request to an advisor by trigger phrase first; if a property journey is active and the request is ambiguous, the orchestrator picks the stage-appropriate advisor.
• State ownership: each advisor writes only its own section of journey.json plus openRisks; the orchestrator is the sole writer of stage and contract; missing values stay null (never invented).
• Coordination loop: route → retrieve property-scoped evidence → run advisor → validate structured result → update owned state → reconcile dependent values → present one next action.
• Guardrails (all agents): property/public records only (never resident dossiers), one property folder per engagement (no cross-property reads), document-traceable figures, and decision support — not legal, financial, lending, insurance, or professional-inspection advice.