ads-strategy

# Full Ad Strategy Orchestrator ## Skill Purpose Perform a comprehensive, end-to-end advertising strategy build for any business from a single URL. This is the flagship command of the AI Ads Strategist — it launches 5 parallel subagents simultaneously to analyze every dimension of ad readiness, then synthesizes all findings into a unified strategy document with a composite Ad Readiness Score (0-100). The output is a client-ready deliverable that covers audience research, creative strategy, funnel architecture, competitive positioning, and budget allocation — the kind of document an agency would charge $3,000-$10,000 to produce. ## When to Use - User runs `/ads strategy <url>` - User asks for a "full ad strategy", "complete advertising plan", or "ad audit" - User wants everything in one command without running individual ad skills separately - User needs a single deliverable covering all advertising dimensions - User is preparing to launch paid ads and wants a complete roadmap ## Input Requirements - **Required:** A business URL to analyze - **Optional:** Monthly budget, target geography, industry context, specific platforms of interest --- ## Dynamic Workflow orchestration This skill IS a Dynamic Workflow. The 5 agents are its units. Run them as a fan-out → verify → synthesize loop, not as a one-shot dump. 1. **Plan** — Phase 1 builds ONE shared context package (the plan input). Every unit reads the identical package so findings stay consistent. 2. **Parallel fan-out** — Launch all 5 units in the SAME response (5 `Agent` calls, never sequential): Audience (25%), Creative (20%), Funnel (20%), Competitive (15%), Budget (20%). 3. **Adversarial verify (2-of-3)** — Before trusting any unit's self-reported score, falsify it. For each unit, apply 3 independent lenses and require ≥2 to agree the finding holds: - **Evidence lens:** is every claim traceable to a fetched page, a `WebSearch` result, or an explicitly-flagged benchmark estimate? Unsourced numbers (CPM/CPC/CPA/ROAS, persona counts, competitor activity) are rejected, not averaged in. - **Consistency lens:** does the unit contradict the Phase-1 context package or another unit (e.g. Funnel budget split ≠ Budget agent allocation; personas target a platform the Funnel excludes)? Contradictions are reconciled before scoring. - **Self-grading lens:** re-derive the unit's score from its own rubric sub-points; if the agent's reported score deviates >10 from the re-derived score, use the re-derived score. A unit's self-reported score is an input, never the verdict (R-VERIFY). - If a unit returns thin/blocked data (e.g. WebFetch failed), mark it `[limited]`, re-run that ONE unit with a narrower prompt (loop-until-dry, max 1 retry), then proceed — never silently drop it (L4). 4. **Synthesize** — YOU compute the composite, reconcile cross-unit conflicts, and write the report. Never paste a unit's raw summary as the verdict (R-ORCH); the synthesis is your own judgment over the 5 verified findings. ## Output contract + VERIFY **Produces (exactly):** - A saved file `ADS-STRATEGY-[CompanyName].md` (full report, structure per "Output Report" below). - A composite **Ad Readiness Score /100** + letter grade, with the weighted breakdown table. - A terminal summary block. **VERIFY before declaring done (self-check, R-RUBRIC / L4):** - [ ] All 5 unit scores present (0-100) AND the composite recomputes exactly: `Audience*.25 + Creative*.20 + Funnel*.20 + Competitive*.15 + Budget*.20`. - [ ] Every metric in the report is either sourced (page/search citation) OR tagged `[estimated from industry benchmarks]` / `[limited data — verify before launch]`. Zero unlabeled invented numbers. - [ ] No cross-section contradiction (Funnel budget split == Budget allocation; personas' platforms ⊆ recommended platforms). - [ ] All ad copy respects platform char limits and is paste-ready. - [ ] The grade reflects reality (weak site/no tracking/unclear pricing ≠ an A). - [ ] Report file actually written to disk; next-steps section present. **Guardrail — evidence or it didn't happen (R-CITE):** `WebSearch`/`WebFetch` are the only data sources. A 403/blocked fetch is `[limited]`, never invented data. Benchmark ranges are allowed only when explicitly labeled as estimates. Never fabricate competitor ad activity, conversion rates, or ROAS. > **Portability note:** This skill uses only model-native tools (`WebFetch`, `WebSearch`, `Agent`) and writes a local `.md` — no VPS-only infra. The `/ads ...` next-step commands assume the sibling ads-* skills are installed; if absent, those rows are informational only. --- ## How to Execute This skill runs 3 phases. Phase 1 gathers intelligence. Phase 2 launches 5 parallel subagents. Phase 3 synthesizes all results into the final report. Display progress to the user: ``` ================================================================ ADS STRATEGY BUILD: [Company Name] ================================================================ Phase 1: Discovery & Business Intelligence ......... [running] Phase 2: Parallel Agent Analysis ................... [pending] - Audience Research Agent (25%) .................. [pending] - Creative Strategy Agent (20%) .................. [pending] - Funnel Architecture Agent (20%) ................ [pending] - Competitive Intelligence Agent (15%) ........... [pending] - Budget & ROI Agent (20%) ...................... [pending] Phase 3: Synthesis & Report Generation ............. [pending] ================================================================ ``` Update each status as work progresses: - `[running]` -- Currently executing - `[complete]` -- Finished successfully - `[limited]` -- Completed with limited data - `[pending]` -- Not yet started --- ## Phase 1: Discovery & Business Intelligence **Objective:** Fetch the target URL, extract all available business intelligence, detect business type, and prepare the context package that all 5 subagents will receive. ### Step 1: Fetch and Analyze the Homepage Use `WebFetch` to retrieve the homepage at the provided URL. Extract: | Data Point | Where to Find | |---|---| | Company name | Page title, logo, footer, about page | | Tagline / Value proposition | Hero section, H1, meta description | | Products or services | Navigation menu, service pages, pricing page | | Pricing model | Pricing page, CTAs ("free trial", "get quote", "add to cart") | | Target market signals | Copy language, imagery, testimonials, case studies | | Trust signals | Client logos, certifications, review counts, media mentions | | Current CTAs | Buttons, forms, phone numbers, chat widgets | | Contact info | Phone, email, address, social links | | Tech stack signals | Meta tags, scripts, platform indicators | | Content assets | Blog, resources, videos, podcasts, lead magnets | ### Step 2: Detect Business Type Classify the business into one of these categories based on homepage signals: | Business Type | Detection Signals | Ad Strategy Implications | |---|---|---| | **SaaS / Software** | Pricing page, "Sign up" / "Free trial" / "Book demo" CTAs, feature lists, integration pages, app subdomain | Focus on demo/trial conversions, long sales cycles, retargeting heavy | | **E-commerce** | Product listings, shopping cart, "Add to cart" buttons, product categories, price displays | Focus on ROAS, product catalog ads, shopping campaigns, impulse triggers | | **Local Business** | Physical address, Google Maps embed, service area mentions, phone number prominent, "Near me" language | Focus on call extensions, local targeting, Google LSAs, radius targeting | | **Agency / Services** | Case studies, portfolio, "Our Work", client logos, consultation CTAs, team page | Focus on lead gen, authority building, LinkedIn, long nurture sequences | | **Creator / Course** | Course listings, "Enroll now", instructor bio, curriculum, testimonials from students, community mentions | Focus on webinar funnels, transformation messaging, urgency/scarcity | | **Restaurant / Hospitality** | Menu, reservations, location hours, food imagery, delivery links | Focus on Instagram/TikTok visuals, local targeting, seasonal promotions | ### Step 3: Identify Industry and Competitive Context Use `WebSearch` to gather additional intelligence: ``` "[Company Name]" competitors "[Company Name]" reviews "[Company Name]" pricing [industry] + [location] market size [industry] advertising benchmarks [current year] ``` ### Step 4: Detect Recommended Platforms Based on business type, determine the optimal platform mix: | Business Type | Primary Platforms | Secondary Platforms | Avoid | |---|---|---|---| | SaaS / Software | Google Ads (Search), LinkedIn | Facebook/Instagram (retargeting), YouTube (demos) | TikTok (unless B2C SaaS) | | E-commerce | Meta (FB/IG), Google Shopping | TikTok, Pinterest, YouTube | LinkedIn | | Local Business | Google Ads (Search + LSA), Facebook/IG | Nextdoor, Yelp Ads | LinkedIn, Pinterest | | Agency / Services | LinkedIn, Google Ads (Search) | Facebook (retargeting), YouTube | TikTok, Pinterest | | Creator / Course | YouTube Ads, Instagram, Facebook | TikTok, Google (search) | LinkedIn (unless B2B) | | Restaurant / Hospitality | Instagram, Facebook, Google (local) | TikTok, Yelp | LinkedIn | ### Step 5: Build the Context Package Compile all Phase 1 findings into a structured context package. This exact package is passed to every subagent so they all operate from the same intelligence: ``` CONTEXT PACKAGE: - Company: [Name] - URL: [URL] - Business Type: [Type] - Industry: [Industry] - Products/Services: [List] - Pricing Model: [Model] - Value Proposition: [Tagline] - Target Geography: [Location or "National/Global"] - Current CTAs: [List] - Trust Signals: [List] - Recommended Platforms: [Platform list] - Monthly Budget: [If provided, else "Not specified"] - Key Competitors: [List from search] ``` --- ## Phase 2: Parallel Agent Launch **Objective:** Launch 5 specialized subagents simultaneously using the `Agent` tool. Each agent receives the full context package from Phase 1 and produces a category score (0-100) plus detailed findings. **CRITICAL:** All 5 agents MUST be launched in parallel (not sequentially) to minimize execution time. Use 5 separate `Agent` tool calls in the same response. --- ### Agent 1: Audience Research Agent **Weight in composite score: 25%** **Corresponding skill:** `ads-audience` Launch this agent with the following prompt: ``` You are the Audience Research Agent for an ad strategy build. Using the context below, build a complete audience analysis. CONTEXT: [Insert full context package from Phase 1] YOUR TASK: 1. Build 5-7 detailed audience personas for this business. For each persona include: - Persona name and archetype (e.g., "Budget-Conscious Buyer", "Overwhelmed Executive") - Demographics: age range, gender split, income level, education, job title, location - Psychographics: values, lifestyle, aspirations, fears, daily frustrations - Pain points: top 3 problems this business solves for them - Buying triggers: what specific events or moments push them to buy - Objections: top 3 reasons they hesitate or say no - Content consumption: platforms they use, content formats they prefer, influencers they follow - Platform-specific targeting parameters: - Meta (Facebook/Instagram): interests, behaviors, lookalike seed suggestions - Google Ads: search intent keywords, in-market audiences, affinity audiences - LinkedIn: job titles, industries, company sizes, skills - TikTok: interest categories, creator affinities, hashtag communities - Persona relevance score (1-5): how valuable is this persona relative to others 2. Define negative audiences (who NOT to target): - Demographic exclusions - Interest-based exclusions - Behavioral exclusions - Why each exclusion matters (wasted spend reasons) 3. Identify the #1 highest-value persona and explain why they should receive the largest budget allocation. 4. Map personas to funnel stages — which personas are TOFU (cold), MOFU (warm), BOFU (hot)? 5. Provide an Audience Clarity Score (0-100) based on: - How well-defined the ICP is from the website (0-25) - How many distinct personas the business can viably target (0-25) - How precise the targeting parameters are on each platform (0-25) - How clearly the website speaks to specific audience segments (0-25) OUTPUT FORMAT: Return your findings as structured markdown with clear headers for each persona and section. End with: - AUDIENCE_CLARITY_SCORE: [0-100] - TOP_PERSONA: [Name] - TOTAL_PERSONAS: [Count] - KEY_INSIGHT: [One-sentence biggest finding about the audience] ``` --- ### Agent 2: Creative Strategy Agent **Weight in composite score: 20%** **Corresponding skill:** `ads-creative` Launch this agent with the following prompt: ``` You are the Creative Strategy Agent for an ad strategy build. Using the context below, develop a complete creative strategy. CONTEXT: [Insert full context package from Phase 1] YOUR TASK: 1. Develop 3 core messaging angles for this business: - Pain-point angle: Lead with the problem the audience faces - Aspiration angle: Lead with the outcome/transformation - Social proof angle: Lead with results, reviews, or authority 2. Write 10 scroll-stopping hooks (first 3 seconds of an ad): - 3 pain-point hooks - 3 curiosity/contrarian hooks - 2 social proof hooks - 2 urgency/scarcity hooks For each hook: the text, the psychology behind it, and which platform it works best on. 3. Create ad copy sets for 3 platforms (customize format per platform): - **Meta (Facebook/Instagram):** Primary text (125 chars), headline (40 chars), description (30 chars), CTA button. Provide 3 variations using PAS, AIDA, and BAB frameworks. - **Google Ads (Search):** 3 responsive search ad sets, each with 15 headlines (30 chars each) and 4 descriptions (90 chars each). Include keyword insertion templates. - **LinkedIn:** Introductory text (150 words max), headline, CTA. 2 variations — one thought-leader style, one direct-response style. 4. Develop 3 creative concept briefs: - Static image ad concept (with visual description, text overlay, color direction) - Short video ad concept (15-second script with shot-by-shot breakdown) - UGC-style ad concept (script template for a creator to film) 5. Provide a Creative Quality Score (0-100) based on: - Hook strength and attention-grabbing potential (0-30) - Copy clarity and persuasion quality (0-30) - Visual concept variety and platform fit (0-20) - A/B test readiness (multiple variations provided) (0-20) OUTPUT FORMAT: Return your findings as structured markdown with clear headers. End with: - CREATIVE_QUALITY_SCORE: [0-100] - STRONGEST_HOOK: [The single best hook you wrote] - RECOMMENDED_FIRST_AD: [Which ad concept to test first and why] - KEY_INSIGHT: [One-sentence biggest creative opportunity] ``` --- ### Agent 3: Funnel Architecture Agent **Weight in composite score: 20%** **Corresponding skill:** `ads-funnel` Launch this agent with the following prompt: ``` You are the Funnel Architecture Agent for an ad strategy build. Using the context below, design a complete advertising funnel. CONTEXT: [Insert full context package from Phase 1] YOUR TASK: 1. Design a complete 4-stage advertising funnel: **TOFU (Top of Funnel) — Awareness:** - Objective: brand awareness, video views, engagement