--- name: whitepaper description: Five-scout team that hunts white-paper topics worth writing for Elite AI Advantage. Each scout works a different signal source and pitches specific named topics with carousel-decomposition built in. Output is a ranked shortlist with a 7-10 slide carousel sketch per top candidate. Sister skill to /prospect (which finds tool ideas) and /forge. trigger: /whitepaper --- # /whitepaper A team of five scouts hunts for **white-paper topics** worth writing for Elite AI Advantage. Each scout works a distinct signal source. Each pitches specific named topics, not themes, not directions, not "you could explore." Named, one-line, with a 7-10 slide carousel sketch built in. After they pitch, you synthesize a ranked shortlist, name the single strongest "write first" pick, kill the duplicates against the existing white-papers index (DB lookup, automatic), and hand the winner off to the writing flow. This is a sister skill to `/prospect`. Prospect goes upstream of `/forge` for tool ideas, this skill goes upstream of the writing flow for long-form authority content. Same five-voice family, calibrated for thesis-grade depth (2,500+ words) and downstream carousel reuse. It is **not** for short blog posts (the TAYA blog pipeline handles those, COST/PROBLEM/REVIEW/EXPLAINER topic shapes). It is **not** for tool ideas (use `/prospect`). It is for the moment when Jake wants a list of authoritative white-paper candidates that decompose cleanly into Instagram/LinkedIn carousel slides for downstream distribution. ## Usage `/whitepaper`, hunt across all signal sources, all funnel stages. `/whitepaper `, focus the hunt. Examples: - `/whitepaper TOFU only`, top-of-funnel SEO-bait, broad-audience papers - `/whitepaper MOFU`, papers that capture intent and feed scoping calls - `/whitepaper for the AI Investment Tier post-launch`, papers that expand the same buyer journey - `/whitepaper that take less than 6 hours to write` - `/whitepaper for mid-market COOs with a board pushing on AI` - `/whitepaper on GEO / answer-engine readiness`, anchor a thesis lane If the constraint genuinely changes the brief (a specific funnel stage, a specific audience, a specific thesis lane), the scouts narrow accordingly. If the constraint is vague ("something good"), ignore it and run the broad hunt. ## What it's for White papers are the long-form authority layer in Jake's content stack. They're the artifacts that get downloaded behind email gates, become source material for blog series and carousel posts, and end up cited in sales conversations as "we already have a paper on that, here's the link." 26 papers already seed the library; new candidates need to be additive, not duplicative, and need to earn their 4-8 hours of writing time. Most "white paper idea" pitches die one of three deaths: 1. **They duplicate something already in the library.** This skill avoids that automatically by querying the WhitePaper table at the start of every run. 2. **They're meandering essays.** Topics that can't decompose into 7-10 distinct slide-able claims fail the carousel test and get killed. 3. **They're vendor-pump pieces.** Generic "ultimate guide to AI" or "5 trends for 2026" content that any consultant could publish and that ages out in 6 months. The five scouts hit different signal sources so the output isn't five flavors of the same brain. Together they cover: 1. **What the buyer is searching for in long-form** (Query Hunter) 2. **Where there's a missing public thesis** (POV Strategist) 3. **Whether the candidate decomposes into a carousel** (Carousel Engineer) 4. **What question the buyer can't answer for their boss** (Buyer's Question) 5. **Which paper feeds the most downstream surfaces** (Anchor Connector) The output is a **ranked shortlist**, 3-5 named topics, each with a one-liner, the signal source that surfaced it, a 7-10 slide carousel sketch, the funnel role, and a flag on which one to write first. Hedging defeats the point. ## What You Must Do When Invoked ### Step 1, Read the brief carefully Before writing anything, identify: - **Is there a constraint?** Funnel stage, audience, thesis lane, write budget? Narrow accordingly. - **What's the implicit ask?** Did Jake just ship something the paper would amplify? Is he low on TOFU traffic? Trying to anchor the GEO / answer-engine thesis with a definitive paper? Sometimes the constraint is in what he just shipped or what's been discussed this session. - **What's already in the library?** This is enforced by the DB lookup in Step 2. Don't skip it. If the brief is genuinely too vague to scout against (e.g. a one-word prompt with no business signal), ask one clarifying sentence. Otherwise proceed. ### Step 2, Pull the existing white-papers index (DEDUP GATE) **Run this exactly once before the scouts work.** The result is the "do not duplicate" list every scout receives. Without it the scouts will pitch papers Jake has already published. ```bash cd "/Users/claude/Documents/ELITE AI ADVANTAGE/_source/eliteaiadvantage" && \ DATABASE_URL="$(railway variables --service Postgres --kv 2>/dev/null | sed -n 's/^DATABASE_PUBLIC_URL=//p')" \ npx tsx -e ' import { PrismaClient } from "@prisma/client"; import { PrismaPg } from "@prisma/adapter-pg"; (async () => { const adapter = new PrismaPg({ connectionString: process.env.DATABASE_URL }); const prisma = new PrismaClient({ adapter }); const papers = await prisma.whitePaper.findMany({ select: { title: true, slug: true, excerpt: true, keywords: true, status: true }, orderBy: { title: "asc" }, }); console.log(JSON.stringify({ count: papers.length, papers }, null, 2)); await prisma.$disconnect(); })(); ' ``` **Note on the IIFE wrapper:** `npx tsx -e` compiles inline to CJS by default, which doesn't support top-level await. The async IIFE is the workaround. If you ever drop the wrapper and the query throws "Top-level await is currently not supported with the cjs output format," that's why. The query intentionally returns title + slug + excerpt + keywords. Title-only matching misses near-duplicates (e.g. "AI Maturity Assessment" vs "Is Your Business AI-Ready?"). The scouts read excerpts and keywords too, and the synthesis enforces the kill on semantic overlap, not just literal title match. **If the DB query fails** (Railway CLI not available, network issue, etc.): say so clearly to Jake and STOP. Do not run the scouts on a stale or empty dedup list, you'll generate duplicates and look like you didn't check. Ask Jake whether to proceed without dedup (rare) or fix the connection first. **Carry the index into every scout's working context.** Each scout's prompt section below references "the existing index", they're expected to reject candidates that overlap before pitching. Synthesis is the second-line filter, not the primary one. ### Step 3, Run the five-scout team Each scout writes **200-350 words, first person, distinct voice**. Each pitches **at least one (max two) specific topics** with: a name, a one-line "what the paper argues," the signal source they're scouting from, the carousel-fitness verdict ("decomposes into N slides"), and the distinguishing element that makes it not a clone of either an existing EAA paper OR a generic competitor publication. No scout is allowed to hedge. No "you could consider…", they each commit to specific named pitches. They can acknowledge other scouts would have different angles, but they don't reference each other by name. Each writes alone. Synthesis is where the voices collide. The five voices: #### 1. The Query Hunter (long-form variant), *search demand for 2,500+ word answers* Listens to what mid-market buyers are typing into Google, Perplexity, and ChatGPT in long-form intent. Different from the short-form Query Hunter in /prospect: the question must be substantial enough that a 500-word blog post couldn't do it justice. Asks: "what 7-12 word search query has high commercial intent and zero authoritative answer that ranks well for it right now?" Pitches papers that occupy a specific search slot. Will not pitch a paper without naming the search behind it. Voice: terse, demand-grounded, allergic to "thought leadership" abstractions. Example pitch shape: *"'How to evaluate an AI consulting proposal', climbing 50%+ year-over-year on Google Trends, no good interactive answer, current SERP is dominated by HubSpot blog posts. Paper: 'The Mid-Market AI Proposal Audit: 12 Red Flags and the Questions That Surface Them.' Carousel-fits as 12 slides, one flag each."* #### 2. The POV Strategist, *missing public thesis* Hunts for places in the AI consulting conversation where there's a thesis nobody will commit to publicly because it offends a vendor, an industry assumption, or an MBA-curriculum talking point. Asks: "what claim does the space NEED someone to publish but won't?" Pitches papers that take a position that ONLY Jake can take credibly because his incentives are clean (no software to sell, no vendor partnerships). Voice: contrarian, sharp, treats "balanced perspective" as a kill signal. Example pitch shape: *"Nobody in the AI consulting space will say publicly that 'AI agents' is a category with 80% rebadged-RPA. The vendors won't because that's their pitch. The agencies won't because they get vendor-referral fees. You can. Paper: 'What AI Agents Actually Are (And What 80% of Vendors Mean When They Say It).' Decomposes into a 9-slide carousel: 1 honest definition, 5 vendor-pitch patterns to spot, 3 questions to ask before signing."* #### 3. The Carousel Engineer, *slide-fitness gate* Tests every other scout's pitches AND surfaces its own candidates pre-filtered for carousel decomposability. Asks: "does this naturally produce 7-10 distinct, slide-able claims?" Topics that survive: list-shaped (5/7/12 of X), contrast-shaped (X vs Y, before vs after), step-shaped (the 3-phase playbook), failure-mode-shaped (where this goes wrong), framework-shaped (named matrix or scorecard). Topics that fail: meandering essays, single-thesis arguments without enumerated supporting structure, anything where the carousel slides would all read like the same slide. Voice: ROI-focused on distribution, suspicious of "deep thoughts" that don't break into pieces. Example pitch shape: *"Paper: 'The 7 Stages of an AI Project That Survives Year One.' Each stage is one slide; carousel writes itself. Paper structure mirrors the carousel exactly, 7 H2 sections, ~350 words each, with a one-line takeaway that becomes the slide headline."* #### 4. The Buyer's Question, *forward-to-boss test* Listens for the actual question the mid-market COO or marketing director is being asked by their CEO/board/CFO that they can't answer. The white paper earns keep when it's the artifact the buyer hands their boss to justify a decision (or to deflect a bad one). Asks: "what question is being asked above the buyer's pay grade that this paper would answer for them?" Pitches papers with a specific reader-handing-this-to-their-boss moment. Voice: empathetic but sharp, names the exact internal-meeting scenario the paper deflects. Example pitch shape: *"COOs are getting asked 'why are we behind on AI' by their boards every quarter and have no shareable artifact that frames the actual answer (which is usually: you're not behind, you're asking the wrong question). Paper: 'Why "Are We Behind on AI?" Is the Wrong Question (And the One You Should Be Asking).' The COO forwards this to the board. 8-slide carousel: 1 reframing, 3 wrong questions, 3 right questions, 1 next-step prompt."* #### 5. The Anchor Connector, *multi-surface reuse* Looks at downstream content surfaces and asks: "which white paper would feed the most of them?" A high-leverage paper feeds: a multi-post blog series, an Instagram carousel, a LinkedIn post pack, a newsletter teaser, sales-deck slides, and a future tool's diagnostic taxonomy. Solo-operator energy budget is the constraint, each paper costs 4-8 hours of writing, so picking topics that yield 5-10 downstream pieces from one source is real leverage. Voice: pragmatic, names the specific downstream artifacts. Example pitch shape: *"Paper: 'The AI Tool Spend Audit: How Mid-Market Teams Are Wasting $40K-200K on AI Subscriptions Without Knowing It.' Feeds: a blog series (3 posts on overlap, dead weight, missing capability), a 9-slide carousel (one waste pattern per slide), a newsletter teaser, a future tool (the AI Stack Diagnosis from prospect's killed list, this paper makes the case the tool serves), and 3 sales-deck objection-handling slides. One paper, 7+ downstream pieces."* #### Format for each scout ``` **The Query Hunter (long-form variant)** [200-350 words, first person, committed pitches with name + one-liner + signal source + carousel-fitness count + distinguishing element] ``` Hard rules: - **Each scout pitches at least one specific named topic.** No "explore the space of…" Named, one-line, buildable into a 2,500+ word paper. - **Each scout names their signal source.** A specific search query, a missing thesis, a downstream artifact pack, a buyer-room scenario. Specific. Receipts. - **Each scout includes the carousel-fitness verdict.** "Decomposes into 9 slides" or "7-step structure, one slide per step." A scout pitching a topic that doesn't decompose has done bad work. - **Each scout explicitly names if a candidate would duplicate something in the existing index.** If yes, kill it before pitching. The dedup list is in their working context. - **No clones across scouts.** If the Query Hunter pitches a "Vendor Decoder" paper and the POV Strategist pitches a "Vendor Honesty Manifesto," synthesis kills the dupe. Each scout pitches from their lane, not converges. - **No platitudes.** "Solves a real problem" is banned. "Drives engagement" is banned. Specific search query, specific carousel structure, specific downstream pack. - **Ground in EAA's actual context**, TAYA, audit, blog pipeline, mid-market positioning, solo-operator constraints, the 26 existing white papers, and any tool/skill in flight (check tools_pipeline.md if it's on the working context). - **No scout references the others by name.** Each writes alone. ### Step 4, Synthesize After all five have pitched, write the synthesis. This is the part Jake takes to the writing flow. ``` --- ## Synthesis **Existing-index dedup check:** [One sentence confirming the DB query returned N papers and no candidate in the shortlist below overlaps semantically. Name any scout-pitched candidate that was killed for overlapping with an existing paper, with the existing paper's title.] **The shortlist:** [3-5 ranked candidates from across the five pitches. Rank by leverage, not by scout count.] For each candidate: **1. **, one-line description of the argument. *Source:* which scout, what specific signal. *Why now:* one sentence on the timing or trigger. *Carousel breakdown (7-10 slides):* 1. [5-word slide headline] 2. [5-word slide headline] ... *Funnel role:* TOFU / MOFU / BOFU + what it captures (email, intent signal, scoping click). *Downstream pack:* what other surfaces this paper feeds (blog series of N posts, X carousel, LinkedIn pack, newsletter, future tool, sales-deck slides). *Write cost:* low / medium / high, how many hours of Jake's writing time this realistically takes. **Write first:** [Pick one. Commit. The single strongest bet, and a one-sentence reason why this beats the others. Leverage = (downstream surfaces × intent capture) / write hours.] **Killed candidates:** [1-3 pitches that didn't survive synthesis. For each: name + one-line kill reason. Standard kill reasons: "duplicates ," "fails carousel decomposition test," "depends on a curated dataset >30 rows," "evergreen-fails (will age out in 6 months)," "no specific buyer-room scenario."] **One concrete next action:** [Specific. "Outline the paper this week, 7 H2 sections matching the carousel slide headlines, 350 words each. Save the outline to the white-papers admin as a DRAFT before drafting." Or: "Validate the search-volume claim for on Google Trends before committing, the Query Hunter cited 50%+ growth, confirm before writing."] ``` ### Step 5, Don't bloat - Total target: ~1,800-2,400 words. Step 2 dedup output (concise) + five scouts × ~250-350 words + ~400-word synthesis. - No introduction, no "great question," no recap of the prompt. - Step 2 should be terse, confirm the count and any dedup hits, then move to scouts. - Start the workshop with `**The Query Hunter (long-form variant)**` and go. - End with the next action. Nothing after. ## Calibration notes for white-paper prospecting The team should know roughly what world Jake operates in: - **Business**: Elite AI Advantage, AI consulting for SMB and mid-market. Solo operator. Premium positioning. TAYA-driven content engine. - **Audience**: AI-skeptical SMB owners and mid-market marketing/operations directors with budget. Tired of consulting jargon. They want answers, dollar ranges, decisions, frameworks they can hand a CFO, not "thought leadership." - **Existing content stack**: - **White papers** (26 seeded; checked at runtime via DB query, DO NOT duplicate these) - The AI Advantage Audit (white paper + interactive /audit tool) - Blog pipeline (TAYA topic types: COST / PROBLEM / REVIEW / EXPLAINER, short-form) - Case studies, skills library - Scope Sketcher (`/scope` MOFU tool, see tools_pipeline.md) - Latest Insights homepage section - **Distribution surfaces** (each candidate paper should feed several): - Blog series (multi-post, TAYA-shaped excerpts of the paper) - Instagram carousels (7-10 slides per paper) - LinkedIn longform + LinkedIn carousel (slide pack reformat) - Newsletter teaser - Sales-deck slide ammo - **Constraints**: Time, not money. Solo-operator energy budget is real. Each paper is 4-8 hours of writing (excluding research). Every public artifact compounds the brand or dilutes it. - **Anti-patterns to flag and kill**: - Generic "ultimate guide to X" papers (clones) - "5 trends for 2026" (ages out in 6 months) - Single-thesis essays that don't decompose into slides - Topics requiring a hand-curated dataset >30 rows (maintenance burden) - Papers that overlap with an existing entry in the DB index - Papers that overlap with the blog pipeline's TAYA shapes (those are short-form territory; white papers earn keep by being LONGER and MORE structured, not just expanded blog posts) - Vendor-pitch-style papers that read like a sales asset (Jake doesn't sell software) When the brief is constrained to a funnel stage: - **TOFU constraint** → Query Hunter + Carousel Engineer carry the load (search demand + maximum reach via carousel reuse) - **MOFU constraint** → POV Strategist + Buyer's Question (the papers that capture intent and feed scoping calls) - **BOFU constraint** → Anchor Connector + Buyer's Question (papers that arm sales conversations directly) When the brief is constrained to a thesis lane (e.g. "GEO" or "AI consulting cost"): - All five scouts narrow to that lane, but still surface from their distinct signal sources within it. Don't let them collapse into five flavors of the same paper. Don't say any of this in the output, let it show in the verdict. ## What to avoid - **No five-voices-pitching-the-same-paper.** If three scouts converge, force the laggards back into their lane. - **No vanity pitches.** "It would be great content" is not enough. "Captures email post-download, feeds 7 downstream pieces, owns the 'AI consulting cost' search slot" is. - **No skipping the dedup query.** If the DB query fails, stop and tell Jake. Don't generate the shortlist and hope none of them are duplicates. Solo operators miss things, that's exactly why this skill exists. - **No "ship it and iterate" mush.** If the team can't commit to a write-first pick, the brief was too vague, re-prompt for a constraint. - **No carousel-fail topics surviving.** A paper that can't be reduced to 7-10 distinct slide headlines doesn't earn its keep here. The Carousel Engineer is the gate. - **No therapy.** "How does this fit your vision" is banned. - **No closing pep talk.** End on the next action. ## Example shape (abbreviated) User: `/whitepaper for the GEO / answer-engine readiness lane` ``` [Step 2 output, abbreviated] Existing index: 26 papers. Two flagged near the GEO lane: - "Answer-Engine Optimization: A Guide for SMB" (slug: answer-engine-optimization-smb) - "How AI Search Changes SEO" (slug: how-ai-search-changes-seo) Both are foundational TOFU primers. Candidates below must offer something new, a deeper diagnostic, a thesis pivot, or a sector-specific angle these don't cover. --- **The Query Hunter (long-form variant)** The query "is my company in ChatGPT" is climbing 80%+ year-over-year and has zero authoritative answer that goes beyond "use Otterly.ai." Adjacent: "GEO checklist mid-market," "answer engine optimization 2026," "ChatGPT brand visibility." Existing index has the foundational primers but nothing diagnostic. Pitch: **"The GEO Visibility Diagnostic: How Mid-Market Brands Are Disappearing from AI Search (And the 9-Step Recovery Playbook)."** Decomposes into a 9-slide carousel, one diagnostic question per slide, then the recovery step. Owns the diagnostic search slot (not just the primer slot). Distinguishes from the existing "AEO Guide for SMB" by being mid-market-specific and prescriptive (the existing paper is foundational + general). [~290 words] **The POV Strategist** [290 words pitching the missing thesis: "The vendors building GEO tools are the same ones who built the broken SEO tools, here's why their tools won't fix this either"] **The Carousel Engineer** [290 words pitching: "The 8 Things Your Content Needs to Be Extracted by AI Engines (Ranked by Effort)"] **The Buyer's Question** [290 words: "Why Your CMO Is Asking About GEO (And the Answer They Actually Need to Take to the CEO)"] **The Anchor Connector** [290 words: "The Mid-Market GEO Playbook: 12-Month Roadmap from Invisibility to Default Citation", feeds the GEO Visibility Check tool from prospect, a 4-post blog series, two carousel packs, sales-deck ammo, and a newsletter sequence. Highest reuse factor] --- ## Synthesis **Existing-index dedup check:** 26 papers in DB. The two GEO-adjacent existing papers ("Answer-Engine Optimization: A Guide for SMB" and "How AI Search Changes SEO") are foundational primers; all five candidates below position deeper or sideways. No semantic duplicates. **The shortlist:** **1. The Mid-Market GEO Playbook: 12-Month Roadmap from Invisibility to Default Citation**, phased roadmap with named diagnostic, fix, and measurement milestones per quarter. *Source:* Anchor Connector, highest downstream reuse, feeds the in-flight GEO Visibility Check tool from /prospect. *Why now:* GEO is the lane that fills the brand hole left by the dropped SEO Audit. A definitive playbook anchors the thesis and gives the GEO tool a "long-form companion" CTA path. *Carousel breakdown (10 slides):* 1. The visibility crisis nobody named 2. Q1: Audit your current AI presence 3. Q1, biggest mistake to avoid 4. Q2: Restructure for extractability 5. Q2, the schema myth 6. Q3: Build citation density 7. Q3, what NOT to do 8. Q4: Measure and defend 9. The 12-month checkpoint 10. Where to start tomorrow *Funnel role:* MOFU. Email-gated download. Captures business + sector. Converts via "book a GEO strategy call" once GEO engagement is productized. *Downstream pack:* 4-post blog series (one per quarter), 2 IG carousels, LinkedIn longform, newsletter sequence, the GEO Visibility Check tool's playbook hand-off, sales-deck objection-handling slides for "why our SEO agency hasn't fixed this." *Write cost:* High (~8 hours). But reuse multiplier brings effective cost per piece below 1 hour each. **2. The GEO Visibility Diagnostic: How Mid-Market Brands Are Disappearing from AI Search (And the 9-Step Recovery Playbook)**, diagnostic-first, prescriptive. *Source:* Query Hunter, owns the diagnostic search slot. *Why now:* climbing search demand + no good interactive answer. Same logic as the GEO Visibility Check tool but in long-form companion shape. *Carousel breakdown (9 slides):* one diagnostic question per slide → recovery step. *Funnel role:* TOFU/MOFU. Captures email + business name. *Downstream pack:* blog series of 3, carousel of 9, sales-deck "this is the diagnostic we run" slide. *Write cost:* Medium (~5 hours). **3. Why Your CMO Is Asking About GEO (And the Answer They Actually Need to Take to the CEO)**, buyer's-room artifact. *Source:* Buyer's Question, exact internal-meeting scenario. *Why now:* CMOs are getting GEO questions in board prep with no shareable answer. *Carousel breakdown (8 slides):* the question, 3 wrong framings, 3 right framings, the next-step prompt. *Funnel role:* MOFU. Captures email + role. *Downstream pack:* LinkedIn longform (CMO audience), 1 carousel, newsletter teaser. *Write cost:* Medium (~4 hours). **Write first:** Run the writing flow on **The Mid-Market GEO Playbook** first. It anchors the entire GEO lane (your strongest brand lane in 2026), feeds the GEO Visibility Check tool's playbook hand-off, and produces the highest downstream piece count of any candidate. The 8-hour write cost is real, but spread across ~10 derivative artifacts it's the lowest effective cost per piece. **Killed candidates:** - *"AI Agents in Plain English"* (POV Strategist's secondary pitch), semantically duplicates "How AI Search Changes SEO" once the AI-agent thread is unpacked. Killed. - *"The 8 Things Your Content Needs to Be Extracted"* (Carousel Engineer), strong concept but partially covered by the diagnostic in #2. Hold; can be a blog post inside the playbook series instead. **One concrete next action:** Outline the GEO Playbook this week, 12 H2 sections (one per month), 300-400 words each, with a one-line "biggest mistake to avoid" pull-quote per quarter. Save the outline as a DRAFT WhitePaper row via /admin/white-papers before drafting the body, so the slug + metadata are locked in and the carousel-script work can run against a stable scaffold. ``` That's the shape. Five scouts, the DB-checked dedup gate, ranked shortlist with carousel breakdowns, and a write-first commit that hands off cleanly to the writing flow.