BEFORE YOU START

You are running the Blog Engine onboarding flow. Your job is to build a complete business profile by analyzing the user's website and finding their direct competitors. Follow these steps exactly and in order.

---

BEFORE YOU START

1. Check for existing business profile

Check if .claude/blog-config.json already exists by attempting to read it.

If it exists and has a non-empty business.business_name:

• Ask: > "A business profile already exists for {business_name} (onboarded on {onboarded_at}). Do you want to overwrite it with a fresh onboarding?"
• Only proceed if they confirm. Otherwise exit.

2. Set up API keys via .env file

Check if a .env file exists in the project root by attempting to read it.

If .env does NOT exist, create it now with this exact content:

# Blog Engine — API Keys
# Fill in the required keys below, then tell Claude you're done.

# Required
DATAFORSEO_LOGIN=your@email.com
DATAFORSEO_PASSWORD=yourpassword
ANTHROPIC_API_KEY=sk-ant-...

# Optional — improve /blog-write article quality
FIRECRAWL_API_KEY=           # better page scraping  (firecrawl.dev — free tier)
TAVILY_API_KEY=              # deeper topic research  (tavily.com — free tier)
YOUTUBE_API_KEY=             # video insights         (console.cloud.google.com)
OPENAI_API_KEY=              # DALL-E article images  (platform.openai.com)

After creating .env, also protect it from accidental git commits. Check if .gitignore exists:

• If it exists and does NOT already contain .env, append .env to it
• If it does NOT exist, create it with just .env

Then tell the user:

"I've created a .env file in your project root and added it to .gitignore so your keys stay private. Open .env and fill in at minimum:

• DATAFORSEO_LOGIN — your DataForSEO email (free account at dataforseo.com)
• DATAFORSEO_PASSWORD — your DataForSEO password
• ANTHROPIC_API_KEY — your key from console.anthropic.com

The optional keys can be added any time before running /blog-write. Reply here when you've saved the file and I'll continue."

Wait for the user to confirm before proceeding.

If .env already exists, read it silently and continue (do not show it to the user).

---

STEP 0 — Get the Website URL

If the user passed a URL as $ARGUMENTS, use that directly. Skip asking.

If no argument was provided, ask:

"What is your website URL? (e.g. https://yoursite.com)"

Wait for their response. Store it as {user_website_url}. Strip trailing slashes.

---

STEP 1 — Scrape the Website

Tell the user: Scraping {user_website_url}...

Use WebFetch to retrieve the website content from {user_website_url}.

Also attempt to fetch these additional pages if they exist (use WebFetch on each, ignore errors silently):

• {user_website_url}/about
• {user_website_url}/pricing
• {user_website_url}/features
• {user_website_url}/product

Concatenate all successfully fetched content into a single block: {scraped_content}.

If the main URL fetch fails entirely, tell the user:

"Could not reach {user_website_url}. Please check the URL and try again." Then stop.

---

STEP 2 — Extract Business Profile

Tell the user: Extracting business profile...

Using the scraped content, produce the business profile JSON by reasoning through the following extraction prompt. Think step by step internally, then output only the final JSON.

Extraction prompt:

You are analyzing a SaaS company's website to extract structured business information for an SEO content strategist.

Below is the website content: <website_content> {scraped_content} </website_content>

Extract the following information. If something is not clearly stated on the website, make a reasonable inference based on context. Do not leave fields empty — use your best judgment and mark inferred fields with "inferred": true.

Return only valid JSON, no explanation, no markdown formatting.

{
  "business_name": "string — company name",
  "website": "string — domain only, no https or www",
  "product_type": "string — one sentence describing what the product is",
  "primary_use_case": "string — the main thing users do with this product",
  "key_features": ["string", "string", "string"],
  "integrations": ["string"],
  "business_model": "string — B2B SaaS / B2C / Marketplace / Agency tool / etc",
  "target_geography": "string — Global / US-focused / specific region if mentioned",
  "pricing_model": "string — subscription / freemium / usage-based / not mentioned",
  "key_differentiator": "string — what makes this product different, in one sentence",
  "icp_signals": {
    "roles": ["string"],
    "industries": ["string"],
    "company_size": "string — SMB / Mid-market / Enterprise / not specified",
    "pain_points": ["string"]
  },
  "brand_voice_signals": "string — describe the tone of the website copy in one sentence",
  "inferred_fields": ["string"]
}

Store the parsed result as {business_profile}.

---

STEP 3A — SERP Competitor Search

Build a search keyword from the business profile:

• Use {business_profile.product_type} as the base
• Format it as: "{product_type} software" (e.g. "email marketing automation software")
• If product_type is vague, fall back to: "best {primary_use_case} tools"

Check for DataForSEO credentials by running:

[ -f .env ] && source .env
echo "LOGIN=${DATAFORSEO_LOGIN:+set} PASSWORD=${DATAFORSEO_PASSWORD:+set}"

If either variable is not printed as set, tell the user:

"DataForSEO credentials are missing or still contain placeholder values in .env. Open .env in your project root, fill in DATAFORSEO_LOGIN and DATAFORSEO_PASSWORD, save, and reply here." Then stop.

Tell the user: Finding competitors via SERP...

Make the following API call using Bash with curl:

[ -f .env ] && source .env
DATAFORSEO_CREDS=$(echo -n "$DATAFORSEO_LOGIN:$DATAFORSEO_PASSWORD" | base64)

curl -s -X POST "https://api.dataforseo.com/v3/serp/google/organic/live/regular" \
  -H "Authorization: Basic $DATAFORSEO_CREDS" \
  -H "Content-Type: application/json" \
  -d '[{
    "keyword": "{search_keyword}",
    "location_code": 2840,
    "language_code": "en",
    "depth": 10
  }]'

Parse the response. Extract the items array from tasks[0].result[0].items.

For each item, keep only: domain, title, description (or snippet).

Store this as {serp_results}. If the API call fails or returns an error status, show the error and stop.

---

STEP 3B — LLM Competitor Filter

Tell the user: Filtering competitors...

Using the SERP results, identify 3 direct competitors by reasoning through the following filter prompt:

Filter prompt:

You are identifying direct competitors for a business from a list of search results.

The business being analyzed:

• Name: {business_profile.business_name}
• Product: {business_profile.product_type}
• Domain: {business_profile.website}

Here are the search results: {serp_results}

Rules:

• Exclude the user's own domain: {business_profile.website}
• Exclude review sites, comparison sites, and aggregators: g2.com, capterra.com, trustpilot.com, getapp.com, softwareadvice.com, producthunt.com, medium.com, reddit.com
• Exclude generic platforms that are not direct competitors: google.com, microsoft.com, apple.com, youtube.com
• Prefer companies whose core product or service directly competes with the analyzed business
• Return exactly 3 competitors. If fewer than 3 clear direct competitors exist, fill remaining slots with the closest indirect competitors from the list
• Return only the root domain without https or www

Return raw JSON only. No markdown. No code blocks. Start with { and end with }.

{
  "competitors": [
    {
      "domain": "string",
      "company_name": "string",
      "reason": "string — one sentence why this is a direct competitor"
    }
  ]
}

Store the result as {competitor_data}.

---

STEP 4 — Save to Config

Get the current timestamp:

date -u +"%Y-%m-%dT%H:%M:%SZ"

Create the .claude/ directory if it doesn't exist:

mkdir -p .claude/exports

Write the following JSON to .claude/blog-config.json:

{
  "version": "1.0",
  "onboarded_at": "{ISO_TIMESTAMP}",
  "business": {business_profile},
  "competitors": {competitor_data.competitors},
  "topics": {
    "last_research_run": null,
    "total_keywords": 0,
    "cluster_count": 0,
    "pipeline": [],
    "used": []
  }
}

---

STEP 5 — Create Scaffold Files

Create two empty scaffold files so the user sees the full structure immediately — before running /blog-topics.

5a — Create CONTENT-PLAN.md in the project root:

Write this file (do not overwrite if it already exists — check first with Read):

# Content Plan — {business_name}

> Managed by Blog Engine · Created: {current_date}
> Run `/blog-topics` to populate this plan with researched keyword opportunities.

---

## Pipeline

Topics queued for writing. Populated after running `/blog-topics`.

| # | Topic | Cluster | Funnel | Vol/mo | KD | Score | Action | Status |
|---|-------|---------|--------|--------|----|-------|--------|--------|
| — | *(run `/blog-topics` to populate)* | | | | | | | |

---

## Written

Articles completed via `/blog-write`. Updated automatically when an article is written.

| # | Topic | Cluster | Funnel | Written | Article URL |
|---|-------|---------|--------|---------|-------------|
| — | *(none yet)* | | | | |

---

## Keyword Clusters

Topic clusters populated after running `/blog-topics`.

*(none yet — run `/blog-topics` to generate clusters)*

---

## Business Profile

- **Business:** {business_name}
- **Product:** {product_type}
- **ICP:** {icp_roles} at {company_size} companies
- **Differentiator:** {key_differentiator}
- **Competitors:** {competitor_1_domain}, {competitor_2_domain}, {competitor_3_domain}

---

## Files

| File | What it contains |
|------|-----------------|
| `CONTENT-PLAN.md` | This file — human-readable pipeline tracker |
| `.claude/exports/` | Keyword CSVs stamped by date — open in Excel or Google Sheets |
| `.claude/blog-keywords.json` | Full keyword data (engine file) |
| `.claude/blog-clusters.json` | Cluster map (engine file) |
| `.claude/blog-config.json` | Pipeline state — do not edit manually |

For {icp_roles}, join business_profile.icp_signals.roles as a comma-separated string. For competitor domains, use the first 3 from {competitor_data.competitors}.

5b — The .claude/exports/ folder is already created in Step 4. No further action needed — CSVs will be written here by /blog-topics.

---

STEP 6 — Confirmation Summary

After saving, show the user a clean summary:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  Onboarding complete — {business_name}
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

  Business Profile
    Product:        {product_type}
    Model:          {business_model}
    ICP:            {roles joined by ", "} at {company_size} companies
    Differentiator: {key_differentiator}

  Competitors Found
    1. {competitor_1.company_name} ({competitor_1.domain})
       {competitor_1.reason}
    2. {competitor_2.company_name} ({competitor_2.domain})
       {competitor_2.reason}
    3. {competitor_3.company_name} ({competitor_3.domain})
       {competitor_3.reason}

  Files Created
    .claude/blog-config.json    business profile + pipeline state
    .claude/exports/            keyword CSVs will be saved here
    CONTENT-PLAN.md             your content tracker — open this anytime

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  NEXT STEP
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

  Run keyword research to populate your content plan:
    /blog-topics         (defaults to US)
    /blog-topics uk      (UK market)
    /blog-topics in      (India market)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

If any fields were inferred from the website, add:

"Note: {N} fields were inferred from context (marked in blog-config.json). Review if needed."