# Local Business Audit

> Local-business prospecting fan-out — audit a city's small businesses (Google Business Profile score, website check) and produce a verified call list / audit sheet with checkpointing so session limits never lose work. Use when the user says "audit <city>", "build a call list for <city>", "find businesses without websites", "GBP audit", or "prospect list".

- Skill: `danielonpurpose/local-business-audit` (Agent Skill)
- Install (CLI): `npx skillmds@latest add danielonpurpose/local-business-audit`
- Raw SKILL.md: https://api.skillmd.com/api/skills/danielonpurpose/local-business-audit/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Security
- Author: DanielOnPurpose (https://skillmd.com/u/danielonpurpose)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/danielonpurpose/local-business-audit

---


# Local Business Audit / GBP Fan-out

Audit a city's small businesses and produce a prioritized call list. Long fan-outs die to session limits — checkpointing is mandatory, not optional.

## The qualifying filter

A prospect is CALL-worthy when: **no website + has a Google Business listing + address in the target city**. Everything else is VERIFY (unclear) or SKIP (has a decent site / outside city / closed).

## Scoring rubric

Score each business 0–10 on digital presence (GBP completeness, reviews, photos, website quality). Keep the rubric fixed across cities so results stay comparable. Read your prior audits before starting a new city so scores stay consistent.

## Checkpointing (MANDATORY)

1. Before fan-out, create a checkpoint dir: `~/Desktop/gbp-audits/<city>-<YYYY-MM>/`.
2. Each agent writes its results to a JSON/markdown file in that dir **immediately on completion** — never hold results only in conversation.
3. Process businesses in batches of ~10–15 per agent; after each batch lands, update `_progress.md` (done / remaining).
4. If a session dies, the resume prompt is: "Continue the <city> audit from the checkpoint dir — read `_progress.md` and audit only the remaining businesses."

## Fan-out procedure

1. Build the business roster for the city (map listings, chamber directories, any available foot-traffic data).
2. Spawn parallel agents per batch: each verifies name, address (right city?), phone, listing link, website (exists? quality?), and scores 0–10 with a one-line justification.
3. Merge into the master sheet: Name | Address | Phone | Listing | Website | Score | Verdict (CALL/VERIFY/SKIP) | Notes.

## Outputs

1. **CSV / sheet** call list (CALL rows first, sorted by score ascending — worst presence = hottest lead).
2. **City summary note** (score distribution, top opportunities, comparison to other cities).