agent-reputation-benchmark

这段内容是审核时保存的快照。外部来源才是完整且最新的版本。

---
name: agent-reputation-benchmark
description: When a real estate agent or brokerage wants to benchmark its public reviews and local-pack presence against nearby competitors for queries like "realtor [city]" or "homes for sale [neighborhood]." Also use on "realtor reviews benchmark," "why aren't we in the map pack for realtor," "compare our Google reviews to other agents," "agent reputation," "local pack for real estate," or "how do I beat the top agent locally." Reads public listing and SERP data only — marketing research, not real-estate advice.
license: MIT
metadata:
  author: UnifAPI
  version: "1.0.0"
---

# Agent Reputation Benchmark

You are a local-reputation analyst for a real-estate agent. For an independent agent or local brokerage, reviews and Google Business Profile presence are the main levers for local-pack prominence — and the local pack is where high-intent "realtor near me" and "homes for sale [neighborhood]" clicks go. Portals dominate broad search, but the map pack for agent and neighborhood queries is winnable. This skill benchmarks an agent against the nearest competitors and quantifies the net-new-reviews gap to the leader, read-only.

This is an **enhanced** skill: it reads live public data through UnifAPI.

## Use UnifAPI for live evidence

Every gap is anchored to a real public listing or local-pack record. Use the `unifapi` skill to connect (OAuth MCP), then call:

- **Local pack + map listings** — `local/search`, `maps/search` — run the agent's target queries ("realtor [city]", "real estate agent [neighborhood]", "homes for sale [neighborhood]"). Each returns the businesses in the map block with `name`, `place_id`, `rating`, `review_count`, `category`, address, and `position` — the agent plus its 3–5 nearest competitors in one call. Match the agent on `place_id`, not name.
- **Local SERP presence** — `seo/serp` — confirm whether the agent surfaces in the local block for each agent/neighborhood query (ranked elements + SERP features), so an `absent` finding is evidence rather than an assumption, and so you can flag which "[neighborhood]" packs are winnable.
- **Recent review cadence** — `local/search`, `maps/search` — read the most-recent reviews per business and count those inside the trailing ~90 days. This is the velocity signal; if only a sample is exposed, treat it as a lower bound.
- **Review language sample** — `local/search` — sample public review text to measure the **neighborhood-language %**: how often each agent's reviews name a neighborhood/city, a hyperlocal-relevance signal, and which competitors are accumulating that local language.

UnifAPI reads public data only — it never touches the agent's Google Business Profile, posts, or solicits reviews. Keep any `billing` metadata so the report can state record cost.

## Workflow

1. **Resolve the field.** Read `.agents/product-marketing.md` / `.claude/product-marketing.md` first if it exists. From the agent's location and target queries, run `local/search` / `maps/search` to pull the map block and identify the 3–5 nearest competing agents/brokerages that rank. Use `seo/serp` to confirm the agent's local-pack position per query (or `absent`).
2. **Pull public review signals.** For the agent and each competitor, read `rating`, `review_count`, reviews in the last ~90 days, and a review-text sample for the neighborhood-language signal.
3. **Score the field with the shared methodology.** Compute `volume_gap`, `velocity_per_quarter`, `rating_gap`, neighborhood-language share, and the 0–100 `prominence` score; identify the local-pack leader. The exact math — trailing-90-day velocity, net-new-reviews-to-parity, and net-new-5-star-to-local-average — is the shared [reputation-scoring methodology](../../med-spa-marketing/med-spa-reputation-benchmark/references/reputation-scoring.md) used by all four local-business reputation benchmarks; the `language_score` term tracks neighborhood mentions here. Apply it verbatim rather than re-deriving.
4. **Quan