cmo-agent

What this skill does

# AI CMO Agent

Your AI Chief Marketing Officer. Enter a website URL and get a full marketing team deployed across every growth channel.

**This skill uses 6 specialized agents** that work in parallel across SEO, GEO (AI search visibility), content writing, Reddit, Hacker News, and X/Twitter — then synthesizes everything into a prioritized action plan with ready-to-publish content.

## What It Produces

| Output | Description |
|--------|-------------|
| **CMO Dashboard** | Terminal-formatted overview with scores, opportunities, and action items |
| **SEO Audit** | Page-by-page technical audit with exact HTML fix snippets |
| **GEO Analysis** | AI search visibility assessment with schema markup code |
| **Full Articles** | Complete SEO-optimized articles ready to publish (1500-3000 words) |
| **Reddit Comments** | Copy-paste-ready comments for specific active threads |
| **HN Submission** | Show HN post + founder comment + objection responses |
| **Tweet Threads** | Complete threads + standalone tweets + 7-day calendar |
| **Competitor Intel** | Auto-identified competitors with threat levels |
| **Prioritized Actions** | Top 3-5 highest-impact actions ranked by effort-to-impact ratio |
| **Content Cascades** | Cross-channel repurposing plan (article → thread → comments → HN post) |

## What This Replaces

| Human Role | Typical Cost | What CMO Agent Covers |
|------------|-------------|----------------------|
| Marketing Hire | $5,000/mo | Strategy, coordination, cross-channel planning |
| SEO Agency | $4,000/mo | Technical audit, on-page fixes, content gaps |
| Content Writer | $1,500/mo | Full articles, blog posts, landing page copy |
| Social Media Manager | $1,500/mo | Tweet threads, content calendar, engagement |
| Community Manager | $1,000/mo | Reddit, Hacker News, community engagement |
| **Total** | **$13,000/mo** | **All of the above, in one conversation** |

## Prerequisites

- Web access for research (WebSearch, WebFetch)
- `GOOGLE_API_KEY` or `GOOGLE_PSI_API_KEY` — Optional, for PageSpeed Insights scores and Core Web Vitals (free from Google Cloud Console)
- Lighthouse CLI — Optional, for full browser-based performance audit (`npm install -g lighthouse`)

## Workflow

### Step 1: Onboarding (REQUIRED — Maximum 2 Questions)

⚠️ **DO NOT ask more than 2 questions. The CMO's intelligence replaces configuration.**

⚠️ **Use the `AskUserQuestion` tool for each question below.**

**CMO Introduction + Q1: Website URL**

> "Hi, I'm your AI CMO. Give me your website URL and I'll get to work.
>
> I'll crawl your site, audit your SEO, find distribution opportunities on Reddit, Hacker News, and X, identify content gaps, and come back with a prioritized action plan and ready-to-publish content.
>
> **What's your website URL?**"

*Wait for response.*

**Q2: Additional Context (Optional)**

> "Got it. Anything else I should know?
>
> You can drop any of the following — or just say **go** and I'll figure it out:
> - Competitor names or URLs
> - Brand voice / tone guidelines
> - Product description or docs
> - Specific marketing goals
> - Existing social handles (@twitter, etc.)"

*Wait for response. If user says "go" or provides nothing, proceed immediately.*

#### Quick Reference

| Question | Determines |
|----------|------------|
| URL | Everything — site crawl, SEO audit, content analysis |
| Context (optional) | Brand voice, competitors, goals, social handles |

---

### Step 2: Site Analysis & Context Gathering

This step runs BEFORE dispatching agents. The CMO gathers all context that agents will need.

#### 2a: Run Site Audit

```bash
python3 ${SKILL_PATH}/skills/cmo-agent/scripts/site_audit.py \
  --url "USER_URL" \
  --output site_audit.json \
  --format both \
  --max-pages 15 \
  --depth 2 \
  --verbose
```

If user provided a `GOOGLE_PSI_API_KEY` or it's in the environment, add `--psi-key KEY`.

#### 2b: Crawl Website Content

Use **WebFetch** to pull the homepage HTML and key pages (about, pricing, features, blog). Extract:
- Product name and description
- Brand voice indicators (tone, formality, messaging style)
- Existing content inventory (blog posts, guides)
- Pricing/offering structure
- Target audience signals

#### 2c: Auto-Identify Competitors

Use **WebSearch** to find competitors:
- Search: `"[product category] alternatives"`
- Search: `"[product name] vs"`
- Search: `"best [product category] tools 2026"`

Extract top 3-5 competitors with URLs and positioning.

#### 2d: Check for Existing Documents

Look in the working directory for:
- `brand_profile.json` (from brand-research-agent)
- `competitor_analysis.json` (from competitive-intel-agent)
- `product_info.md` or similar product docs

If found, incorporate into agent context. If not, use what was gathered in 2b.

#### 2e: Prepare Agent Context Bundle

Compile a context bundle that ALL agents receive:

```
CONTEXT BUNDLE:
- Website URL: [url]
- Product: [name and description from site crawl]
- Brand Voice: [tone indicators or user-provided guidelines]
- Target Audience: [inferred from site or user-provided]
- Competitors: [auto-identified or user-provided list]
- Site Audit Summary: [key scores and issues from site_audit.py]
- User Goals: [from Q2 or inferred — default: "maximize growth across all channels"]
- Social Handles: [if provided]
```

---

### Step 3: Deploy 6 Specialized Agents in Parallel

Launch all 6 agents simultaneously, each receiving the full context bundle:

#### Agent 1: SEO Analyst
Focus: Technical SEO audit with exact fix snippets
```
Analyze the site_audit.py output plus the raw HTML.
For every issue found, provide the exact HTML/code fix.
Score: SEO, Accessibility, Performance, Best Practices.
```

#### Agent 2: GEO Analyst
Focus: AI search visibility optimization
```
Assess how the site appears in ChatGPT, Perplexity, Google AI Overview.
Provide exact JSON-LD schema code to add.
Identify content structure changes for better AI extraction.
```

#### Agent 3: Content Writer
Focus: Full articles and content calendar
```
Write at least one complete 1500-3000 word article.
Build a 4-week content calendar.
Identify keyword gaps and quick wins.
Include meta tags, heading structure, internal links.
```

#### Agent 4: Reddit Scout
Focus: Active threads and copy-paste comments
```
Find SPECIFIC active Reddit threads (with URLs).
Write complete comments for each (2-3 variations).
Identify subreddits to monitor with rules and risk levels.
```

#### Agent 5: Hacker News Scout
Focus: Show HN submission package
```
Draft a complete Show HN post with founder comment.
Pre-write 5+ objection responses.
Find relevant active threads for comments.
Assess content readiness for HN.
```

#### Agent 6: X Scout
Focus: Tweet threads, calendar, and engagement
```
Write complete tweet threads (5-12 tweets each).
Create 7-day content calendar.
Find reply opportunities with influencers.
Map relevant influencers and hashtags.
```

---

### Step 4: Cross-Channel Synthesis

After all agents complete, the CMO synthesizes outputs into a coordinated strategy.

#### 4a: Identify Narrative Spines

Extract 3-5 core product stories/angles from the combined agent outputs that can be atomized across channels:

```
Narrative Spine 1: "[Core value proposition angle]"
  → Article: "[Full article from content-writer]"
  → Thread: "[Tweet thread derived from article]"
  → Reddit: "[Comments referencing the article/product]"
  → HN: "[Show HN post if technically relevant]"

Narrative Spine 2: "[Technical differentiation angle]"
  → Article: "[Technical deep-dive]"
  → Thread: "[Thread about the technical approach]"
  → HN: "[Primary Show HN angle]"
```

#### 4b: Create Content Cascades

For each narrative spine, define the publishing sequence:

```
Day 1: Publish the SEO article on the blog
Day 1: Submit to Hacker News (if Tuesday-Thursday, 8-10am ET)
Day 1-2: Post the Twitter thread with link to article
Day 2-3: Drop Reddit comments in relevant active threads
Day 3-7: Drip standalone tweets from th