SEO Agency Killer — Claude Code Skill (April 2026)

Philosophy: One command in, real Google data out. No browsing, no scraping,
no guesswork — every metric below comes straight from DataForSEO's live APIs.

Phase 0: First-Run Setup Check (MANDATORY — run before ANY command)

Before doing anything with a /seo ... request, run this check. Do not skip it, even if the user just wants a quick command.

Step 1 — Read the credentials file:

cat ~/.claude/skills/seo/.env 2>/dev/null

Step 2 — Decide if credentials are configured:

The .env is NOT configured if any of these are true:

• The file doesn't exist
• DATAFORSEOLOGIN is missing or contains the literal string yourloginemailhere
• DATAFORSEOPASSWORD is missing or contains the literal string yourapipasswordhere

Step 3a — If credentials ARE configured: continue to the requested command, no message needed.

Step 3b — If credentials are NOT configured: STOP. Do not run the requested command. Instead, paste exactly this setup wizard to the user:

🔧 DataForSEO API Setup Required

This skill needs DataForSEO API credentials to fetch real Google SEO data.
One-time setup, takes 60 seconds:

  Step 1 — Sign up free (includes $1 trial credit, ~10 full audits):
           https://l.dataforseo.com/4epETnH

  Step 2 — After verifying your email, open API Access:
           https://app.dataforseo.com/api-access

  Step 3 — Copy your API Login (your email) and API Password
           (the long alphanumeric string under "API password")

  Step 4 — Paste both back to me in this exact format:

           login: [email protected]
           password: 1adc9025dc1a3e86

I'll save them securely to ~/.claude/skills/seo/.env (chmod 600) and run
your original command immediately after.

Then WAIT for the user's reply with credentials.

Step 4 — When the user pastes credentials:

1. Parse login: and password: from their message (handle Login:, Password:,

email-only, or just two lines as well — be lenient)

1. Write them to ~/.claude/skills/seo/.env using this format:

DATAFORSEO_LOGIN=<their_login>
DATAFORSEO_PASSWORD=<their_password>

1. Lock down permissions:

chmod 600 ~/.claude/skills/seo/.env

1. Run a verification call (cheap — ~$0.001):

~/.claude/skills/seo/scripts/keyword_research.py volume "test" 2>&1

1. Interpret the result:

Important: Never echo the credentials back to the user, never include them in any tool output, and never commit them to git.

Quick Reference

Setup

1. Sign up at https://l.dataforseo.com/4epETnH (free trial includes $1 credit).
2. Add credentials to ~/.claude/skills/seo/.env:

DATAFORSEO_LOGIN=your_login_email
DATAFORSEO_PASSWORD=your_api_password

The installer creates this file from .env.example and the shared client at [scripts/dataforseoclient.py](scripts/dataforseoclient.py) reads it.

How To Run Each Command

All scripts below live at ~/.claude/skills/seo/scripts/ and have shebangs pinned to the skill's isolated venv. Call them directly — no python3 prefix.

/seo quick <domain>

A 60-second snapshot. Run these three calls in parallel and synthesize:

~/.claude/skills/seo/scripts/domain_overview.py overview --target <domain>
~/.claude/skills/seo/scripts/backlinks.py summary --target <domain>
~/.claude/skills/seo/scripts/domain_overview.py ranked --target <domain> --limit 25

Produce a 5-line summary: estimated organic traffic, total keywords ranking, backlink count, top 5 keywords by volume, and one headline opportunity.

/seo prospect <niche> <city> — Find SEO clients

~/.claude/skills/seo/scripts/prospect_finder.py find \
    --niche "<niche>" \
    --city "<city + state>" \
    --limit 30 \
    --out ~/.claude/skills/seo/output/<niche>-<city>-prospects.json

Returns 30 local businesses ranked by closeability score (0-100) — the ones most likely to need (and pay for) SEO help. See the dedicated seo-prospect sub-skill for full output formatting.

/seo email <domain> — Draft cold outreach

Reads the saved audit JSON for the domain and drafts a personalized cold email that references specific findings (current ranking, missing meta tags, estimated traffic value of a single opportunity). No new API calls — pure synthesis from the audit data. See the dedicated seo-email sub-skill.

Pre-requisite: /seo audit <domain> must have run first.

/seo audit <domain> — Full Audit (Parallel Subagents)

Delegate to the five specialist subagents in parallel using the Agent tool:

Each subagent calls the appropriate scripts, summarizes findings, and returns a structured block. Then:

1. Compute the composite SEO Score as a weighted average:

overall = 0.25keywords + 0.25technical + 0.20competitors + 0.15content + 0.15*authority

1. Build an audit JSON object matching schema/audit_input.example.json.
2. Write it to ~/.claude/skills/seo/output/<domain>-audit.json.
3. Offer to generate the PDF: /seo report-pdf <domain>.

/seo keywords <seed>

~/.claude/skills/seo/scripts/keyword_research.py related "<seed>" --limit 200
~/.claude/skills/seo/scripts/keyword_research.py suggestions "<seed>" --limit 100

Group results by intent (informational, commercial, transactional, navigational). Surface the 20 best opportunities ranked by volume / (difficulty + 1). Always include CPC so the user sees commercial value.

/seo technical <domain>

~/.claude/skills/seo/scripts/on_page_audit.py site --target <domain> --max-crawl-pages 100

This kicks off a real crawl and waits for results (can take 2-5 minutes for 100 pages). When complete, group issues by severity (critical, high, medium, low) and produce a prioritized fix list. Always include estimated impact.

For a quick single-page audit:

~/.claude/skills/seo/scripts/on_page_audit.py page --url <url>

/seo competitors <domain>

~/.claude/skills/seo/scripts/domain_overview.py competitors --target <domain> --limit 20

For the top 5 competitors, also pull their ranked keywords and intersect with the user's domain to identify SERP overlap and content gaps. Output:

• Top 10 competitors table (domain, organic keywords, traffic estimate)
• Shared keywords vs. unique keywords for each competitor
• 3 specific competitors to study most closely (and why)

/seo content <domain>

~/.claude/skills/seo/scripts/domain_overview.py ranked --target <domain> --limit 200

Cluster the ranked keywords into topics. Identify:

• Strong topics (5+ keywords ranking top 10)
• Weak topics (lots of keywords, none top 10 — opportunity)
• Missing topics (competitors have them, you don't — use content_gap)

/seo backlinks <domain>

~/.claude/skills/seo/scripts/backlinks.py summary --target <domain>
~/.claude/skills/seo/scripts/backlinks.py refdomains --target <domain> --limit 50
~/.claude/skills/seo/scripts/backlinks.py anchors --target <domain> --limit 30

Report on: total backlinks, referring domains, dofollow ratio, top anchors (flag over-optimization), top referring domains by rank, and any toxicity indicators (spam score, suspicious patterns).

/seo rankings <domain> <keyword1> <keyword2> ...

~/.claude/skills/seo/scripts/serp_check.py rank --domain <domain> --keywords kw1 kw2 ...

Returns position (or "not in top 100") for each keyword. Group by:

• 🟢 Position 1-3 (winning)
• 🟡 Position 4-10 (page 1 — fight for top 3)
• 🟠 Position 11-30 (close — push to page 1)
• 🔴 Position 31-100 / not ranking (long-haul or pivot)

/seo content-gap <you> <competitor>

~/.claude/skills/seo/scripts/domain_overview.py content_gap --you <you> --competitors <comp>

Returns keywords the competitor ranks for and you don't. Sort by search volume × ranking position (closer to page 1 = easier wins for them).

/seo compare <domain1> <domain2>

Run overview, ranked (limit 50), and summary for both domains in parallel, plus an intersect call to find shared keywords. Output a side-by-side comparison table covering keywords, traffic, backlinks, referring domains, top shared keywords, and a verdict.

/seo report-pdf <domain>

~/.claude/skills/seo/scripts/generate_pdf_report.py \
    --input ~/.claude/skills/seo/output/<domain>-audit.json \
    --output ~/.claude/skills/seo/output/<domain>-report.pdf

Requires that /seo audit ran first and saved the JSON. Open the PDF when done.

Output Conventions

• Save all intermediate JSON to ~/.claude/skills/seo/output/<domain>-<command>.json
• Always print a tight executive summary first (≤8 lines), then the detail
• Use score badges: 🟢 80+, 🔵 60-79, 🟡 40-59, 🔴 <40
• End every audit with a "Top 3 Actions This Week" section

DataForSEO API Notes

• All scripts use the live endpoint variants (synchronous, no polling)

except on_page/site which requires a true crawl

• Default location is United States, language en — pass --location /

--language to override (e.g. --location "United Kingdom" --language en)

• A typical full audit costs ~$0.10-0.30 in DataForSEO credits
• The shared client is at [scripts/dataforseoclient.py](scripts/dataforseoclient.py)

When NOT To Use This Skill

• AI search visibility (ChatGPT, Perplexity, Gemini citations) → use /geo instead
• Google Business Profile / local pack tracking → DataForSEO has it but a

dedicated local-SEO tool would be a better fit

• Pure content writing (no data needed) → use a content skill directly