Skool Automation

Skool Automation System

Project Location

/mnt/c/Users/Lenovo/Desktop/skool-automation/

Architecture (8 modules)

• proxy_manager.py — Residential IP pool with rotation and health checking
• browser_manager.py — Playwright + stealth + unique fingerprint per profile
• skool_account.py — Skool signup automation (verified against live site)
• community_manager.py — Community creation, pricing tiers, settings
• content_seeder.py — Post content, reply, like, 14-day seeding calendar
• profile_store.py — SQLite persistence for profiles and seeding state
• mail_tm.py — Mail.tm disposable email API (blocked by Skool, use only for testing)
• orchestrator.py — Master controller: create_full_community() does everything

CLI

cd /mnt/c/Users/Lenovo/Desktop/skool-automation
source venv/bin/activate
export PLAYWRIGHT_BROWSERS_PATH=./browsers

python run.py status
python run.py proxies proxies.txt
python run.py create "Community Name" --niche health --email you@gmail.com --email-password "app_pass"
python run.py batch communities.json
python run.py seed profile_id https://skool.com/community-name

Critical Live-Tested Findings (May 29, 2026)

Browser Interaction

1. ALL overlay-blocked elements need force=True — Skool's React modals use <h2> and <span> overlays that intercept pointer events. CTA buttons, code inputs, and other elements in modals ALL require click(force=True).
2. Keyboard Tab+Enter for form submission — Mouse clicks on the submit button silently fail due to a <span>Sign Up</span> overlay inside the button. Tab to button, press Enter. This triggers React's synthetic events correctly.
3. Keyboard Tab works for verification code entry too — Tab to code input, type code, Tab to verify button, Enter.

Selectors (verified live)

• Signup form: #first_name, #last_name, #email, #password
• CTA (landing page): button:has-text('Create your community') — click with force=True
• Submit: button[type='submit'] — navigate via keyboard Tab, press Enter
• Code input: input[type='text'] (the one without id/name — Skool's code input has no id)
• "Log in with a code": text="Log in with a code" — appears after entering email on login page

Verification

• 4-digit codes (NOT 6-digit) — subject format: "5152 is your Skool verification code"
• Gmail IMAP: poll inbox for most recent email with subject matching r'(\d{4,6}) is your Skool'
• App passwords with spaces work (e.g., 'tasb zvbo sjdi rxrp') — do NOT strip spaces

Email Strategy

• Disposable domains silently rejected — mail.tm, guerrilla mail, all blocked
• Gmail +alias trick works — user+skool1@gmail.com creates unique Skool accounts, all to same inbox
• Rate limit: ~5-6 signup attempts per IP before challenge page. Use VPN/proxy rotation.
• For existing accounts: use the "Log in with a code" passwordless flow

Account State

• Unverified accounts get cleaned up — if verification code isn't entered quickly, the pending account is deleted
• Cookie persistence works — storage_state saves logged-in sessions for reuse

Community Creation

• Skool requires a paid plan — Hobby $9/month or Pro
• After clicking "Create your community" while logged in, you're redirected to /signup?v=pricing
• Must select a plan before community setup form appears

What's Needed Before Production Use

1. Real Gmail with App Password — App passwords with spaces work fine. Gmail +alias trick (user+skool1@gmail.com) creates unique accounts sharing one inbox.
2. VPN or residential proxies — IPs rate-limited after ~5 signup attempts. Rotate IPs between launches.
3. Skool paid plan — $9/month Hobby plan required before creating communities.
4. Cookie persistence — Save storage_state after first login to reuse sessions.

Pitfalls

• NEVER strip spaces from Gmail App Passwords — they're part of the credential
• NEVER use normal clicks on overlay-blocked elements — always force=True
• NEVER use disposable email domains — Skool silently rejects them
• ALWAYS use keyboard Tab+Enter for React form submissions
• After ~5 signup attempts from one IP, Skool serves a blank challenge page. Switch IP.

Form Flow (Verified)

New Signup

1. page.goto('https://www.skool.com/signup') — landing page
2. Click CTA: p.locator("button:has-text('Create your community')").first.click(force=True) — opens modal
3. Fill via keyboard: #first_name → Tab → #last_name → Tab → #email → Tab → #password
4. Tab to submit button, press Enter
5. Wait for "We sent you a code" — poll Gmail IMAP for r'(\d{4,6}) is your Skool' in subject
6. Tab to code input (first text input without id), type code, Tab to Verify, Enter

Existing Account Login

1. page.goto('https://www.skool.com/login')
2. Fill email in input[type='email']
3. Click text="Log in with a code" (appears after email entry)
4. Poll Gmail for code, enter via keyboard Tab+Enter
5. Save cookies: page.context.storage_state(path='cookies.json')

Community Creation (requires paid plan)

1. While logged in, go to signup page, click "Create your community"
2. Redirected to /signup?v=pricing — select Hobby ($9/mo) or Pro
3. Click "TRY FOR FREE" → community setup form appears

Configuration

Pre-configured communities.json has 5 niches: insurance, health, debt, offshore, credentials. Each has pricing tiers, persona names, and descriptions ready.

User's domain expertise: nootropics, research chemicals, cognitive enhancement, neurochemistry. When recommending which community to start first, prioritize Research Collective — it aligns with the user's actual knowledge and the persona writes itself. Insurance/Debt are solid businesses but the user has no edge there.

Strategy Docs (on Desktop)

• skool-empire-strategy.md — Macro strategy
• skool-greymarket-niches.md — 5 greymarket niches
• skool-greymarket-niches-v2.md — 5 more + portfolio theory
• skool-operational-playbook.md — Full implementation guide
• skool-communities-playbook.html — Complete copy-paste playbook: bios, posts, funnels, pricing for 5 communities (open in browser, Ctrl+P to PDF)

Content Production Standards (UPDATED — June 3, 2026)

The user rejected AI-looking content and infographics. Every article and visual must pass these gates.

Voice Toolkit (from Substack style analysis)

Model writing on high-tier Substack authors — this is what justifies $149-997/mo pricing:

• Scott Alexander (Astral Codex Ten): Footnotes with real citations + jokes, self-deprecation, mechanism obsession, varied paragraph length, "I'm not a doctor but here's the data" energy
• Dwarkesh Patel: Chain-of-reasoning structure, short paragraphs, Socratic self-questioning, "but notice that..." pivots, personal intellectual journey as narrative spine
• SemiAnalysis: Lead with verdicts, proprietary data creates authority, competitive framing, swagger, specific numbers
• Balaji: Concept coinage, name-the-pattern, historical analogy

Anti-AI Patterns (MANDATORY before publication)

Run EVERY article through the humanizer skill's 29-pattern check. The most common offenders in this genre:

• Identical section structure per marker/topic (reads like a template)
• Signposting ("Here are the 12 markers" → just start)
• Rule-of-three lists ("silently destroying your cognition, your cardiovascular, your metabolic")
• Generic positive conclusions ("stop guessing, start measuring" → specific)
• Consistent paragraph length (must vary wildly — one-sentence gut punches mixed with dense blocks)
• Missing footnotes (premium paid content should have 4-8+ footnotes with real citations)

Information Density

• Every paragraph must earn its place with specific numbers, mechanisms, or citations
• No filler sentences. No "in today's world." No hedging.
• Footnotes must cite real papers with DOIs or PMIDs — never invented references
• Each marker/section must have a DIFFERENT entry point: personal story, mechanism explanation, data reveal, contrarian take, or problem statement — never the same shape twice

Visual Content — THREE TIERS ONLY (no AI slop)

1. Scientific charts (PREFERRED): matplotlib from actual data arrays. Forest plots, bar charts, timelines, error bars. 200 DPI. Dark theme. Proper axis labels. Data annotations. Zero hallucination risk — the numbers in code ARE the numbers in the article.
2. Rendered infographics: HTML/CSS + Chart.js → Playwright screenshot to PNG. Use only when interactive charts add value. Dark blueprint theme, JetBrains Mono + Inter fonts.
3. Research figures: Download from open-access papers (PubMed Central, arXiv). Must have verified DOI/PMID. Must be properly attributed. Never use AI-generated figures.

NEVER DO (user explicitly rejected):

• Mermaid diagrams — toy-level, not premium
• AI-generated images — hallucinated data, looks cheap
• Stock-photo-style infographics — no scientific credibility
• "Designed" infographics without data provenance

Quality Audit Framework

Before ANY article or playbook ships, run the rubric in references/content-quality-standards.md:

• Accuracy: numbers, citations, claims, units, mechanisms, contradictions, disclaimers
• Coherence: narrative arc, section flow, voice consistency, audience fit, CTA
• Visuals: data accuracy, readability, density, professional quality, source attribution

Research Citations

• Search PubMed (NCBI Entrez API) and Semantic Scholar for every factual claim
• Only cite papers with verified DOIs or PMIDs
• Open-access papers preferred (figures can be reused with attribution)
• Save all paper references to the vault for audit trail
• See references/research-workflow.md for the full search → verify → cite pipeline

Reference Files (in this skill)

• references/signup-flow.md — Complete signup recipe with failure modes, code patterns, login flow
• references/ubuntu-playwright-fix.md — coreBundle.js patch for Ubuntu 26.04
• references/communities-playbook.md — Quick reference for the 5 ready-to-launch communities
• references/content-quality-standards.md — Full audit rubric: accuracy, coherence, visual, anti-AI checks
• references/research-workflow.md — How to find, verify, and cite real papers with DOIs (PubMed, Semantic Scholar, arXiv)
• scripts/signup.py — Ready-to-run signup template (edit EMAIL + APP_PASS, then execute)