// Work on EstateWise MCP, agentic-ai, web-grounding, token flows, and A2A integrations. Use when changing tool behavior, orchestration, runtime flags, MCP wrappers, HTTP/A2A endpoints, or AI integration paths.
Extend or audit Codex support in the EstateWise repository. Use when changing .codex config, .codex/rules, .agents/skills, AGENTS.md layering, multi-agent roles, project guidance discovery, or Codex-specific documentation and workflows.
Audit producer-consumer alignment in EstateWise when REST, tRPC, MCP, gRPC, A2A, or shared payloads may have changed. Use when one package's change could silently break another package, tool, or doc set.
Execute production-safe code changes in the EstateWise monorepo. Use when implementing or fixing behavior in backend, frontend, MCP, agentic-ai, gRPC, deployment-control, tests, or repo docs. Do not use for pure brainstorming, skill installation, or work outside this repository.
Run or debug the EstateWise local environment. Use when setting up services, choosing the smallest service combination, checking local endpoints, or troubleshooting environment/config issues. Do not use for routine code implementation unless local run/debugging is the main task.
Work on EstateWise deployment-control, Docker, Kubernetes, Helm, Terraform, Jenkins, and cloud deployment assets. Use for infrastructure and operational workflow changes, not ordinary application feature work.
Review EstateWise changes for correctness, regressions, security issues, contract mismatches, operational risk, and missing tests. Use for PR review, diff review, bug hunts, and branch-vs-main analysis. Do not use for implementation tasks unless the user asks for fixes after findings are reported.
| name | estatewise-ai-runtime |
| description | Work on EstateWise MCP, agentic-ai, web-grounding, token flows, and A2A integrations. Use when changing tool behavior, orchestration, runtime flags, MCP wrappers, HTTP/A2A endpoints, or AI integration paths. |
Use this skill when the task touches the tool-first AI stack rather than just the app UI/API.
backend/: source APIs behind property, graph, analytics, auth, finance, and chat behavior.mcp/: stdio server that exposes those capabilities as tools.agentic-ai/: consumers and orchestrators that call MCP tools, plus HTTP/A2A access.mcp/src/server.tsmcp/src/core/config.tsmcp/src/tools/index.tsmcp/src/tools/a2a.tsmcp/src/tools/system.tsmcp/src/tools/monitoring.tsagentic-ai/src/index.tsagentic-ai/src/http/server.tsagentic-ai/src/orchestrator/AgentOrchestrator.tsagentic-ai/src/lang/tools.tsagentic-ai/src/lang/graph.tsagentic-ai/src/a2a/cd mcp
npm run build
npm run client:dev
npm run client:call -- <tool> '<json>'
Remember: npm run dev waits for a stdio client and appears idle by design.
cd mcp && npm run build
cd ../agentic-ai && npm run build
npm run dev "Find 3-bed homes in Chapel Hill and compare two ZPIDs"
cd agentic-ai
npm run build
npm run dev -- --langgraph "Compare two homes and estimate mortgage"
Use only when explicitly touching the Python runtime. Requires agentic-ai/crewai/requirements.txt and OPENAI_API_KEY.
Key endpoints in agentic-ai/src/http/server.ts:
GET /healthPOST /runGET /run/streamGET /.well-known/agent-card.jsonGET /a2a/agent-cardPOST /a2aIf these change, inspect mcp/src/tools/a2a.ts immediately.
agentic-ai and MCP bridge tools.client:call fails because a tool name or schema changed.agentic-ai/src/lang/tools.ts still expects the old MCP payload shape.mcp/src/tools/a2a.ts still targets an older endpoint or response structure.THREAD_ID behavior changes without docs/examples being updated.mcp/README.mdagentic-ai/README.mdREADME.mdARCHITECTURE.mdRAG_SYSTEM.md when AI-system behavior materially changes