AdFast — Knowledge Base

The "brain" of the tool. The audit/plan/execute engine reads these packs + research files at runtime (delivered license-gated from the hosted MCP server). Nothing here calls an LLM by itself; the host agent (the user's own Claude Code / Codex / opencode token) reasons over this material.

Structure

adfast/
├── packs/                      # machine-readable knowledge (the auditor's rulebook)
│   ├── viral-product-principles.json   (Marc Lou 31 principles)         32 rules
│   ├── hormozi-offers.json             (Value Equation, Grand Slam)     11 fw / 30 rules
│   ├── hormozi-leads.json              (Core Four, lead magnets)        13 fw / 35 rules
│   ├── hormozi-money-models.json       (attraction→continuity)          11 fw / 29 rules
│   ├── hormozi-advertising.json        (hooks, ad anatomy, angles)      14 fw / 29 rules
│   ├── marclou-marketing.json          (headline spec, free-plan math)   8 fw / 27 rules
│   ├── channel-playbooks.json          (8 channels, 75 decision rules)
│   └── _fragments/                     (per-channel source objects)
└── research/                   # human-readable depth (the strategist's library)
    ├── teardowns/              (5 live ShipFast-ecosystem site teardowns)
    ├── channels/              (8 channel deep-dives with sources + confidence)
    ├── marclou-growth-essays.md
    └── reddit-channels.md
├── skills/                     # design EXECUTION skills (HOW the tool changes code)
│   ├── 00-brand-safety.md      (mandatory pre-edit scan + guardrails)
│   ├── landing-page.md · pricing-page.md · animation.md
└── config/
    └── design.config.json      (palette / motion + assets policy: ask | html-css | hybrid)

Totals: 57 frameworks + 182 rules across 6 content packs, 8 channel playbooks (75 decision rules), 15 research documents.

How the tool consumes it (pack → audit phase)

| Phase / sub-agent | Reads | |---|---| | Page Auditor | viral-product-principles, marclou-marketing, hormozi-offers (hero/headline/copy/cta/proof/visual rules) | | Pricing Analyst | hormozi-money-models, hormozi-offers (pricing/offer/guarantee), marclou-marketing (free-plan math) | | Angle Strategist | hormozi-offers (Value Equation), hormozi-advertising (angle library), viral-content, marclou-growth-essays | | Competitor Scout | channel-playbooks, hormozi-leads (Core Four), teardowns (patterns-to-steal) | | Planner | everything, ranked by impact/effort; routes each fix to its automation bucket |

Automation buckets (honesty contract, enforced in every rule)

• auto — pure copy/layout/CTA/structure the tool may implement directly (with git checkpoint).
• advisory — pricing / business-model / channel / strategy calls. The tool recommends; the founder executes (e.g. Stripe). Never auto-applied.
• needs_user_input — anything requiring a real truth (testimonials, metrics, founder story/video, real scarcity/deadline). The tool reserves the slot and asks; it never fabricates proof.

CHANNEL DECISION MATRIX

The core of the "which marketing channel for you" engine. Read top-to-bottom: stage and budget gate the channel; audience picks within it.

The hard truth that shapes everything

Every paid channel has a real floor of roughly $1,500–3,000/mo (the algorithms need ~50 conversions/ad-set/week to exit the learning phase). A $500 paid test almost always returns a false "it doesn't work." So for broke / pre-PMF founders the correct default is not paid. This matches the Marc Lou + Hormozi thesis exactly: earn distribution before you buy it.

Budget ladder (× stage)

| Budget / stage | Primary channels | Notes | |---|---|---| | $0, time only — pre-PMF | Organic + community (platform by audience), build-in-public, launch-everywhere (PH + 25 directories), referrals/word-of-mouth, founder UGC, SEO if demand exists | Referrals are the #1 first-customer channel (40/326 projects). Cost is 5–20 hrs/wk for 6–18 mo. No paid. | | <$500 — early | All of the above + nano/micro influencer (gifting/affiliate only) + retargeting-only paid (if you already have traffic) + free-tool / programmatic SEO | Cold paid acquisition does not work at this budget. | | $500–1.5k — early, some traction | + Micro-influencer roster (5–15, flat+CPA hybrid, whitelist winners) + UGC content production ($50–500/video) feeding organic & Spark | UGC is a creative input, not a standalone channel; it needs paid/owned distribution. | | $1.5–3k — proven funnel + tracking | + ONE cold paid channel, mastered: Meta (B2C visual, AOV<$80) or TikTok (B2C impulse, <35, weekly creative) or Google (existing search demand, B2B/high-intent) | Pick one. Don't spread a thin budget across three. | | $3k+ — scaling | Multi-channel paid + influencer whitelisting + scale winners + SEO as the compounding long game | Now paid is a multiplier on a proven economic engine (LTV:CAC ≥ 3:1). |

Audience overlay (picks the channel within a budget tier)

| Audience | Strong | Weak / skip | |---|---|---| | B2C visual / impulse | TikTok, Reels/IG, Meta, UGC, micro-influencer, viral short-form | Google (unless real search demand) | | B2B SaaS | Founder-led LinkedIn/X content, cold outreach (founder-led +30–50% reply), Google Search (intent), niche-creator trust | TikTok/Meta cold (retarget only) | | Dev tool / technical | X, Hacker News, Reddit (value-first), DevHunt, SEO, build-in-public | Most paid social (wasted) |

Cross-cutting rules (true at every tier)

1. Capture email everywhere — the #1 missing piece. follower→email ~1–3%, email→customer ~2–6%, and email converts ~4x better than any other source.
2. SEO is a parallel 6–12 month long-game, not a launch channel. Never promise traction before ~3 months. Programmatic SEO + a free tool is the Marc Lou engine for demand-rich niches.
3. Creative/hook is ~65% of paid performance. Hook = headline. The same hook frameworks power ads, organic, and the landing hero.
4. Recommend behaviors with confidence; treat outcomes as survivor-biased. Never promise "viral."

Cross-cutting findings from the 5 site teardowns

• Uniform, cheap tech stack: all five are Next.js on Vercel + Tailwind/DaisyUI (some shadcn). Animation is framer-motion whileInView + IntersectionObserver plus occasional custom requestAnimationFrame. No Rive/GSAP/Lottie anywhere. The "viral indie" look is a reusable, low-cost pattern, not heavy animation.
• Proof-as-product flywheel: the proof IS the marketing (TrustMRR's FAKE-vs-REAL OG chart and live leaderboard; DataFast's product-as-demo iframe; SuperShrimp's XP leaderboard).
• Offer/pricing range to copy from: $17 one-time + pain-cost anchor (SuperShrimp) · high-ticket + rising seat-ladder + anti-guarantee + loss-aversion (Ship-or-Die $299) · course + discount + live countdown + clickable proof (CodeFast) · event-based subscription, no scarcity (DataFast) · free + sponsor/marketplace (TrustMRR).
• Free spinoff tools → SEO/distribution loop (DataFast's DataCity/DataRadar; Marc's 7-free-tools-a-year play).

Status & next step

Knowledge base: complete and validated. Next: scaffold the hosted MCP server skeleton (phase state machine + tool stubs + license/pack delivery) and wire these packs in as the served IP.