| name | riley-brown-marketing-automation |
| description | Riley Brown's (Agent Native / Chorus) agentic marketing-operations doctrine — running a company's entire marketing through a coding agent + a portfolio of small, named, composable skills wrapped around scraper APIs. Examples-over-instructions (retrieve verified exemplars, don't prompt harder), creator→skill voice compilation, longest-running-ad competitor intel, template-steal ad factories, email drafts at scale, constraint-encoded booking links, "turn it into a skill" freezing, and per-task model/cost routing — all terminating in a human-editable draft behind approval. Use for scraping creators/competitors, building marketing skill libraries, ad-spy, on-brand creative from winners, inbox/scheduling ops, and agent-native distribution. |
Riley Brown — Agentic Marketing Operations
The gap this fills: the roster has content geniuses (Kallaway, Lara, Cole), ad experts (Dara, Luke), and system builders (Saraev, Kashef) — but nobody whose domain is "the agent IS the marketing department." Riley owns the operations tier: scraping, distribution, scheduling, inbox, file hygiene — the unglamorous layer that makes the content tier compound. Core truth: AI can't verify content quality the way it verifies code, so the job is feeding the agent verified examples on demand, judging with real taste, and freezing every capability into a named skill that compounds.
Expert: Riley Brown (@rileybrownai), AI-native founder of Chorus (open agent platform) and Vibecode. Runs his startup's marketing entirely inside Codex + custom skills. Source: 3 verified Riley Brown videos + a 100-frame visual analysis of the primary demo (extractions/riley-brown/). Extraction: extractions/riley-brown/mes-extraction.md.
Load order: Tier 1 = this file + the workflow. Tier 2 (+genius.md) for any composition, creative, or client-facing work. Verbatim prompts + claims ledger: references/source-quotes.md.
Costs — Our Routes vs. Riley's Stack
Riley demos a paid third-party stack (ScrapeCreators, Foreplay, Firecrawl, Buffer, Paper). We hold none of those keys and don't need them. Every workflow below routes through infrastructure we already own — mostly $0, with a small metered Apify budget for social scraping. Riley's original tools are noted only as what he uses, never as a dependency.
| Capability | Riley's tool (paid) | Our route ($0 unless noted) |
|---|
| Scrape a creator | ScrapeCreators | /scrape-creator → social_intel.py (Apify sc-* actors ~$0.005–0.25/run; yt-dlp captions free) |
| Competitor ad-spy | Foreplay ($59–149/mo) | /ad-spy → Meta Ad Library via read-only Playwright, $0 |
| Winning-ad → creative | Paper.design | /creative-from-winners → Dara / Fantastic Studio / Canva / Higgsfield |
| Brand-asset scrape | Firecrawl | /brand-asset-scrape → Tavily + Playwright, $0 |
| Email drafts at scale | Gmail | /inbox-drafts → Gmail MCP (drafts only), $0 |
| Social scheduling | Buffer/Typefully | /post-scheduler → Typefully (key pending) |
| Booking links | Cal.com | /scheduling-links → Cal.com API (key pending) |
| Data warehouse | Notion | Social Intelligence DB (3a749875-a897-8104-a867-fc9aeb53f52c) via notion_api.py |
Full mapping: references/api-integration-guide.md.
Workflow Table (12 workflows, 3 tiers)
Tier 1 — Foundation (the core moves)
| Workflow | Produces | Use when |
|---|
workflows/riley-scrape-to-skill.md (front door) | Creator → named voice-writing skill, deploy-in-voice on a fresh topic | Turning any creator into a callable style |
workflows/riley-ad-spy.md | Longest-running competitor ads, ranked, inference-labeled analysis | Competitive creative intel |
workflows/riley-turn-it-into-a-skill.md | Any successful run frozen as a named, inspectable skill/workflow | A task will recur; a correction should stick |
workflows/riley-inbox-drafts.md | N voice-matched email drafts, never sent | Inbox backlog; decline-and-pitch at scale (his highest-leverage move) |
Tier 2 — Practitioner (the ops + creative layer)
| Workflow | Produces | Use when |
|---|
workflows/riley-creator-analyzer.md | Per-post "why it works" verdicts grounded in a hook lens | A scraped corpus needs the surpass-Riley analysis pass |
workflows/riley-template-steal-ads.md | On-brand variant batch off a winning ad's structure | Volume ad testing from a proven skeleton |
workflows/riley-brand-scrape.md | Structured brand-asset sheet (colors/type/logos/voice) | Grounding creative in a real brand |
workflows/riley-distribution-ops.md | Staged social posts + constraint-encoded booking links + file hygiene | The unglamorous distribution/scheduling/booking tier |
Tier 3 — Stacking (cross-expert pipelines)
| Workflow | Pairs with | Use when |
|---|
workflows/riley-parallax-pipeline.md | Farrice voice + Satori + Parallax | Research → scrape → essay → design → distribute under Farrice's brand |
workflows/riley-lara-amplifier.md | lara-acosta-, diandra-escobar- | Scraped LinkedIn corpus → ghostwritten posts |
workflows/riley-dara-adfactory.md | dara-static-engine (Meg/Luke as copy lens) | Ad-spy intel → original static-ad production |
workflows/riley-automations.md | any workflow above | Promoting a useful one-off to a recurring/scheduled automation ("act in the future") |
Quick Reference (the spine)
DIAGNOSE (verification gap) → OWN THE TASTE → SCRAPE VERIFIED EXEMPLARS → TURN IT INTO A SKILL → CHAIN → DRAFT-LINK TERMINUS → DIAL MODEL/COST → CORRECT INTO THE FILE / AUTOMATE
- Content is subjective and unverifiable — supply examples, don't prompt harder.
- Only delegate what you can judge; taste is the load-bearing human input.
- Exclude sponsored/boosted posts with retained exclusion evidence.
- Longest-running ad = the winner, but labeled inference from durability, not ROAS proof.
- The durable asset is a named skill, born from a successful run — read what it wrote.
- Skills chain live; pick MCP / REST / computer-use per tool.
- Never auto-send. Every action ends in an editable draft/link behind approval.
- Per-task model + effort dial; open-source for mechanical work.
- Write corrections into the skill file so they compound; promote useful one-offs to automations.
Stacking Guide (consistent with extractions/riley-brown/vision.md)
- Riley × Kallaway — he literally scrapes Kallaway on camera.
/scrape-creator → /extract deepens voice replication; skills/kallaway-* grounds the "why it works" verdicts.
- Riley × Dara Denney — longest-running-ad intel from
/ad-spy feeds /dara-static-engine production (riley-dara-adfactory).
- Riley × Lara / Diandra — scraped LinkedIn corpora → hook/voice engines (
riley-lara-amplifier).
- Riley × Meg Heckman / buyer-trigger — scraped competitor ads → trigger audits.
- Riley × the extraction pipeline itself — any scraped creator becomes an
/extract candidate (scrape → corpus → skill is our own loop, industrialized; /scrape-creator marks Extract Candidate).
- Farrice deployments — VOICE-CARD + dial is a mandatory layer on anything shipped under his name (binding
farrice_voice_alignment).
Anti-Patterns (reject on sight)
Prompt-engineering a voice from scratch · scraping raw engagement without excluding sponsored · presenting ad duration as ROAS proof · auto-sending/auto-posting with no editable-draft terminus · cloning a competitor ad word-for-word or carrying its real byline (the "Dr. Fahim Hussain" failure) · fresh prompt instead of a saved named skill · corrections left in chat · delegating in a domain you can't judge · inventing performance numbers the source doesn't expose · treating a skill as a black box instead of reading the file it wrote.
Execution Prompts (structure-pure v2)
9 deterministic practitioner prompts — each carries an Output Contract, Output Skeleton, and Quality Gate. When a deliverable matches one, Read it and honor its contract instead of improvising the output shape.
- Batch Email Drafts — [INTENT] —
skills/riley-brown-marketing-automation/references/prompts-v2/batch-email-drafts.md
- Brand Assets — [BRAND DOMAIN] —
skills/riley-brown-marketing-automation/references/prompts-v2/brand-asset-sheet.md
- Competitor Ad Intel — [COMPETITOR SET] —
skills/riley-brown-marketing-automation/references/prompts-v2/competitor-ad-intel-report.md
- Voice-Skill: [SKILL NAME] — [CREATOR] —
skills/riley-brown-marketing-automation/references/prompts-v2/creator-voice-skill-deploy.md
- Why-It-Works Analysis — [BATCH TAG] —
skills/riley-brown-marketing-automation/references/prompts-v2/creator-why-it-works-analysis.md
- Distribution Staging — [DATE] —
skills/riley-brown-marketing-automation/references/prompts-v2/distribution-staging-package.md
- Durable Asset — [SKILL NAME] —
skills/riley-brown-marketing-automation/references/prompts-v2/durable-asset-forge.md
- Exemplar-Grounded Voice Pipeline — [PIPELINE TARGET] —
skills/riley-brown-marketing-automation/references/prompts-v2/exemplar-grounded-voice-pipeline.md
- On-Brand Ad Variant Batch — Source: [SOURCE AD] —
skills/riley-brown-marketing-automation/references/prompts-v2/on-brand-ad-variant-batch.md
