Growisto Seo

Growisto SEO Potential — Claude Code Plugin

Two skills for SEO outreach potential analysis. Built so any analyst with Claude Code + an Ahrefs subscription can run the same analysis Growisto SEO consultants do manually — without needing the standalone Streamlit dashboard.

Why a plugin instead of a dashboard?

• No LLM API costs — Claude (in your Claude Code chat) does the reasoning
• Portable — clone, install, ready to go
• Composable — chain with other Claude Code skills

---

What's inside

seo-plugins/
├── .claude-plugin/plugin.json
├── commands/
│   ├── seo-find-competitors.md     →  /seo-find-competitors <url>
│   └── seo-gap-analysis.md         →  /seo-gap-analysis --client-name X --client-csv ...
└── skills/
    ├── find-competitors/           →  Skill 1: URL → ranked competitor list
    └── gap-analysis/               →  Skill 2: Keyword CSVs → SEO potential report

Skill 1 — find-competitors

Input a client URL. The script fetches:

1. Ahrefs organic-competitors (top 15 by keyword overlap)
2. Ahrefs site-metrics for the client (traffic + KW count)
3. Optional: scans homepages of client + each candidate for category context
4. Sitemap-based category seeds (e.g. /diamond/, /rings/ for ORRA)

It writes JSON. Then Claude reads it and applies positioning judgement — picking the 3 most category-aligned competitors and suggesting brands the analyst should add manually.

Cost: ~260 Ahrefs units per run.

Skill 2 — gap-analysis

Input client + competitor Ahrefs keyword CSVs. The script:

1. Filters non-brand commercial+transactional intent keywords
2. Drops blog/help/resources URLs
3. Computes gap keywords (where competitors rank top 10, client doesn't)
4. Sorts by competitor's actual captured traffic
5. Computes "Big Wins" (top 3 by score) + achievable target
6. Detects competitor mis-alignment

It writes JSON. Claude reads it, decides which keywords are in client's catalog scope, then the script regenerates a Word report (report-{client}.docx) using only the in-scope set.

Cost: $0 — all reasoning happens in Claude chat, no API calls.

---

Installation

Option A — point Claude Code at this folder directly

# Clone the parent repo (this plugin lives inside it)
git clone https://github.com/SagarShanmathuran/growisto-seo-agent.git
cd growisto-seo-agent

# Tell Claude Code to load the plugin
mkdir -p ~/.claude/plugins
ln -s "$(pwd)/seo-plugins" ~/.claude/plugins/growisto-seo

Then in Claude Code:

• /plugin reload
• The /seo-find-competitors and /seo-gap-analysis commands appear

Option B — copy the folder into your existing ~/.claude/plugins/ directory

cp -r seo-plugins ~/.claude/plugins/growisto-seo

---

Configuration

Create seo-plugins/.env (gitignored — never commit):

AHREFS_API_TOKEN=your-ahrefs-api-token-from-https://ahrefs.com/api/profile
SEARCHAPI_KEY=your-searchapi-key-optional

Only AHREFS_API_TOKEN is required. Get it from your Ahrefs subscription's API tab.

The script also reads from a .env in the parent repo root if present — so if you've already configured the dashboard version, this just works.

---

Usage

1. Find competitors

/seo-find-competitors https://locobuzz.com/ --country IN --scan

Claude will:

• Run the script (~260 Ahrefs units)
• Read the JSON
• Apply positioning reasoning
• Present a markdown table with top 3 picks + suggested manual additions

2. Pull Ahrefs CSVs

For the picks Claude recommended, manually export from your Ahrefs UI:

• Site Explorer → Organic keywords → Export CSV
• Repeat for client + each competitor

3. Run the gap analysis

/seo-gap-analysis \
    --client-name Locobuzz \
    --client-url https://locobuzz.com/ \
    --client-csv ~/Downloads/locobuzz-keywords.csv \
    --competitor "Sproutsocial:~/Downloads/sproutsocial-keywords.csv" \
    --competitor "Hootsuite:~/Downloads/hootsuite-keywords.csv" \
    --competitor "Brand24:~/Downloads/brand24-keywords.csv" \
    --business-model b2b_saas \
    --niche "social listening platform"

Claude will:

• Run the script (no API cost)
• Read the candidate keywords
• Decide which are in scope (their actual analyst-style call)
• Re-run with --finalize to generate the DOCX
• Summarise in chat with verdict + Big Wins + warnings

The DOCX lands at /tmp/report-{slug}.docx.

---

How this differs from the dashboard

Aspect	Streamlit dashboard	Plugin
Where it runs	A web app on Streamlit Cloud	Inside your Claude Code chat
Who's the LLM	Gemini API + Claude API as fallback	Claude (your chat session)
LLM cost	Free tier 429s common, paid Claude as fallback	$0 — your existing Claude subscription
Multi-user	Yes (URL + password)	One analyst at a time
Deployment	git push → auto-deploy	git pull → manual reload
Best for	Multi-analyst team workflow	Solo analyst, ad-hoc analyses

You can use both. The dashboard for batch screening; the plugin when you want Claude's reasoning involved interactively.

---

Troubleshooting

"AHREFS_API_TOKEN not set" — create seo-plugins/.env (or the repo-root .env) with the token.

Script can't find pandas / requests / python-docx — run pip install pandas requests python-docx beautifulsoup4 lxml (matches the dashboard's requirements.txt).

CSV format errors — Ahrefs exports two formats: comparison (UTF-16 tab-separated, has "Previous/Current" columns) and snapshot (UTF-8 comma-separated). The loader handles both. If your CSV has different column names, paste a header sample and we'll add the alias.

Claude doesn't see the commands — /plugin list should show growisto-seo. If not, re-check the symlink/copy path and run /plugin reload.