# Orbid Match

> Use when the user wants to match an Orbid catalog product to a tender, run_match, 对标, 匹配产品, 用目录配一下, or score a named model against a session. Requires connected Orbid MCP. Not for filling Excel, Go/No-Go as a tool, or listing gaps.

- Skill: `leon-sg/orbid-match` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add leon-sg/orbid-match`
- Raw SKILL.md: https://api.skillmd.com/api/skills/leon-sg/orbid-match/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Data & Analytics
- License: MIT
- Author: Leon-SG (https://skillmd.com/u/leon-sg)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/leon-sg/orbid-match

---


# Orbid match

Match **one catalog product** to an analyzed tender. MCP never fills Excel. `run_match` spends match credits.

## Gate

Need Orbid MCP tools `list_sessions`, `search_product_catalog`, `query_structured_specs`, `run_match` (names may be prefixed). Missing / 401 → stop. Connect `https://app.orbid.dev/mcp`. Do not invent scores.

## Steps

1. Resolve `sessionId` from `list_sessions` (name search if they named the bid). Required. Do not invent ids. Do not pass org_id, workspaceId, or skipAutoFill.
2. Resolve product: `search_product_catalog` with a **category** keyword (ventilator, infusion pump, ultrasound). Pick the row’s **UUID** `productId` and exact `productName`. A model string is not an id.
3. Optional: `query_structured_specs` `mode=lookup` with that `productId` before matching, if they asked about numbers/certs.
4. `run_match({ sessionId, productId, productName })`. Omit `forceRerun` unless they asked to re-run. Report JSON: fit, deviations, credits. Say Excel was **not** filled; confirm-before-fill stays in the Orbid bid session.

If the host only connected `/mcp/readonly`, `run_match` is absent — stop and say they need the full connector URL.

## Do not

- Call `run_fill_excel` or claim the xlsx was written.
- Pass a model name as `productId`.
- Match without `sessionId` + UUID + exact name.