Competitor Analysis v2

Purpose: An 8-phase competitive analysis of a client plus up to 5 competitors — website intelligence, optional 22-attribute ad scoring, automated screenshots, gap identification, and mandatory client verification — producing a strategic Client Gift (10-15 page PDF), a tactical Ads Angle Brief (1 page), or both.

Credentials Required: None. The whole workflow runs on Playwright plus web fetches. Competitor ad copy for the optional ad-scoring phase comes from a SERP tool or is pasted in — no Google Ads API access needed.

This skill deliberately does NOT:

• See competitor spend, impression share, or auction data — public surfaces only; decline asks that need them
• Put Google Ads mechanics in the Client Gift, or hand the Ads Brief to a client
• Ship a claim without client-website verification — Phase 6 gates both outputs
• Build the final RSAs — the brief's 30-char angle sketches feed ad-copy-generation-framework and rsa-single-account
• Monitor competitors over time — every run is a dated snapshot

When to Use This Skill

Ask the user at the start which output they want (for white-label partner work, "Both" is the usual answer; for a relationship gift, "Client Gift" alone). The choice decides which prompts run — the run table is in references/workflow-contract.md § Which prompts run, per output.

The 8-Phase Workflow

The seven files in prompts/ are the methodology authority for their stages — this SKILL routes to them and never restates their scoring criteria. Four of them carry internal phase labels from the framework's v1 numbering; translate via the contract's numbering table rather than trusting a prompt's label against the phases below.

Phase 1: Gather Inputs

- Client name + URL
- Industry/market
- 5 competitor URLs (or request search assistance)
- Output type: "Client Gift" | "Ads Brief" | "Both"

Competitor selection is a judgment call with its own rules — who earns one of the five slots, what to exclude, why an honest 3 beats a padded 5: rules.md § Which competitors earn a slot. If the user has no list: WebSearch 8-10 candidates, present with brief descriptions, get approval on 5.

Phase 2: Capture Screenshots (MANDATORY)

node .claude/skills/competitor-analysis-v2/scripts/screenshot.cjs \
  https://client.com \
  https://competitor1.com https://competitor2.com https://competitor3.com \
  https://competitor4.com https://competitor5.com \
  --output /tmp/competitor-report/screenshots

Client URL first — index 1 is the client everywhere downstream. Full URLs only (a schemeless URL is silently dropped). Check the Screenshots captured: X/Y line, not the exit code. Prerequisite: npm install playwright. Capture mechanics, retry behavior, and the screenshots.json metadata contract: references/workflow-contract.md § screenshot.cjs.

Phase 3: Extract Website Content (Parallel)

Launch 6 parallel Task agents, one per URL (client + 5 competitors):

- subagent_type: "general-purpose"
- prompt: "Use WebFetch to analyze [URL]. Use this extraction prompt:
  [paste from prompts/website-extraction.md]. Return only the XML output."

Serial fetching takes 5-6 minutes; parallel completes in ~1. Output: one 21-field XML profile per site (schema owned by prompts/website-extraction.md). An empty field on a JS-heavy site can be a fetch artifact — cross-check the screenshot before recording it as a gap (rules.md § Screenshot reads).

Phase 4: Analyze Competitor Ads (If Available)

If SERP results or pasted ads exist: run prompts/strategic-analysis.md (attributes 1-15, 45 points) and prompts/tactical-scan.md (attributes 16-22, 21 points) per competitor ad, then feed scores to Phase 5. If no ads: skip, note "Ad analysis not performed (no competitor ads provided)" — website analysis is sufficient for the Client Gift.

Phase 5: Gap Identification

Run prompts/gap-identification.md — messaging matrix, positioning map, emotional landscape, strategic/tactical gap tests (table stakes vs differentiators vs whitespace). Blank scaffolds live in prompts/analysis-framework.md. This phase identifies opportunities, never recommendations — that distinction is the prompt's own critical rule.

Phase 6: Client Verification (MANDATORY)

Run prompts/client-verification.md: scrape the client site, build the verified capabilities inventory, map every gap to ✅ can-fill / ❌ cannot-verify with source citations. No recommendation survives without verification — this phase gates both outputs and is non-negotiable. It is the ad-copy-verification-standard in action.

Phase 7: Generate Output(s)

• Client Gift (10-15 pages, for the CLIENT, zero Google Ads mechanics) — structure owned by templates/client-gift.md (15 sections, cover through summary). Tone: strategic, insight-driven, no jargon.
• Ads Angle Brief (1 page, for the PRACTITIONER) — structure owned by templates/ads-angle-brief.md (7 sections: positioning axis through ad-to-page rules). Angle development runs prompts/angle-development.md on verified gaps only; angle-choice judgment (verification strength beats gap value, compliance demotions, small-sample phrasing) is rules.md § Angle selection.

Phase 8: PDF Generation (For Client Gift)

Build HTML with embedded CSS from templates/template.css, place screenshots, render positioning maps as CSS grids, then:

const { chromium } = require('playwright');
(async () => {
  const browser = await chromium.launch();
  const page = await browser.newPage();
  await page.goto('file://' + process.cwd() + '/[client]-competitor-report.html', { waitUntil: 'networkidle' });
  await page.pdf({
    path: '[client]-competitor-report.pdf',
    format: 'A4',
    printBackground: true,
    margin: { top: '0', right: '0', bottom: '0', left: '0' }
  });
  await browser.close();
})();

Quality Checklist

Before delivery:

• All 6 screenshots captured successfully (X/Y line, screenshots.json)
• Website content extracted for all 6 URLs (thin fetches cross-checked)
• Client verification completed (Phase 6)
• All recommendations have source citations
• Client Gift has NO Google Ads mechanics
• Ads Brief uses ONLY verified angles; "Cannot Recommend" list intact
• Positioning maps are accurate
• Property management client → Fair Housing check on every angle

Files in This Skill

After a Run

Nothing lands in the skill folder. The run's footprint, all in your chosen output location: the screenshot PNGs plus screenshots.json (ISO capturedAt, per-URL success/retried — the record of what ran and what's missing), and the deliverables ([client]-competitor-report.html → .pdf, the brief's markdown). Date the deliverables; every run is a snapshot, and the next one starts fresh.

When to Load Other Skills

Example Usage

"Run competitor analysis for Smith Auto Repair" → ask which output → Both → client URL + 5 competitor URLs (or search flow) → full 8-phase run → Client Gift PDF + Ads Brief markdown.

"Create a client gift for Example Property" → Client Gift only → phases 1-3, 5-8 (Phase 4 skipped unless ads are provided) → strategic PDF, delivered with the property-management framing from sales/service-page-copy.md.

Installation

mkdir -p .claude/skills/competitor-analysis-v2/scripts \
         .claude/skills/competitor-analysis-v2/prompts \
         .claude/skills/competitor-analysis-v2/templates \
         .claude/skills/competitor-analysis-v2/sales \
         .claude/skills/competitor-analysis-v2/references
for f in SKILL.md README.md rules.md examples.md \
         references/workflow-contract.md scripts/screenshot.cjs \
         prompts/website-extraction.md prompts/strategic-analysis.md \
         prompts/tactical-scan.md prompts/gap-identification.md \
         prompts/client-verification.md prompts/angle-development.md \
         prompts/analysis-framework.md templates/client-gift.md \
         templates/ads-angle-brief.md templates/template.css \
         sales/service-page-copy.md; do
  curl -o ".claude/skills/competitor-analysis-v2/$f" \
    "https://raw.githubusercontent.com/fourteenwm/ppc-ai-skills/main/competitor-analysis-v2/$f"
done

Version History

• v2.0 (2026-01) - Merged the 22-attribute ad analysis framework and the strategic competitor report into a single 8-phase workflow
• v1.0 (2025-12) - Original 22-attribute ad copy analysis framework