Jia Jingqiu

Open-source Markdown

README.md and skill file.

This page mirrors the public repository Markdown so reviewers can inspect the workflow logic without leaving the portfolio site.

README.md

# B2B Account Radar AI Kit

Category-based B2B account discovery and account-research planner for
cross-border commerce.

This repo is built for finding customer opportunities, not just sending email:

```text
category / market
  -> search query plan
  -> company account scoring
  -> social surface mapping
  -> buyer-role hypothesis
  -> outreach angle
  -> CRM next action
```

It can involve LinkedIn, Facebook, directories and company websites, but it does
not scrape logged-in pages, bypass platform rules, or auto-DM people. Feed it
lawful public observations, manually exported lead lists, official API output or
directory data.

## What It Does

- Generates public search queries by category, market, buyer role and channel.
- Scores B2B accounts by category fit, market, buyer signal, website quality,
  social surface and content gap.
- Maps LinkedIn/Facebook/Instagram/TikTok/website URLs as research surfaces.
- Produces a relationship-first next action instead of mass email.
- Creates channel-specific talking points for sourcing, AI content support or
  independent-site marketing.

## Quick Start

```bash
python3 -m venv .venv
source .venv/bin/activate
python -m pip install -e .
account-radar query-plan examples/sample_market.json
account-radar score examples/sample_accounts.json examples/sample_market.json
```

Run tests:

```bash
python -m unittest discover -s tests
```

## Open Source Reference Map

This repository is original code. It references:

- [Crawlee](https://github.com/apify/crawlee): public web data collection
  architecture inspiration.
- [Scrapy](https://github.com/scrapy/scrapy): crawler/spider design patterns.
- [Crawl4AI](https://github.com/unclecode/crawl4ai): LLM-friendly public page
  extraction inspiration.
- [SearXNG](https://github.com/searxng/searxng): metasearch workflow inspiration.
- [Chatwoot](https://github.com/chatwoot/chatwoot): conversation routing after
  inbound or consent-based contact.

## Compliance Position

Use this repo for account research and planning. Do not use it for credentialed
scraping, hidden API access, spam, fake identities, or automated social DMs.

## License

MIT.

skills/account-radar.skill.md

# account-radar.skill

Purpose: find and qualify B2B customer accounts by category and market without
turning the workflow into scraping or spam.

## Loop

```text
category + market
  -> public query plan
  -> account evidence collection
  -> account score
  -> buyer-role hypothesis
  -> social surface review
  -> official channel or warm interaction
  -> CRM memory
```

## Output

- Search queries for website, LinkedIn, Facebook and directories.
- Account score and reasons.
- Social surface map.
- Next relationship-first action.

## Rules

- Do not scrape logged-in LinkedIn or Facebook.
- Do not auto-DM.
- Use public pages, directories, official APIs or manually exported lists.
- Record source URL and reason for fit.
- Separate account research from message sending.