For Claude Code users who write their own skills
You already write skills. You know the problem: every skill picks up its own voice, every output reads slightly off-brand, and updating positioning means rewriting half your folder. This pack solves that with one master-prompt.md that every skill reads.
Battle-tested in 4 production brands across PL/EN/DE markets. Closed an €8K deal from the connected pipeline last month. MIT license. See architecture →
See packages →If you've shipped 10+ skills, you've hit this
Most skill packs on Gumroad solve "I need a skill that does X." That's a tools collection. Useful, but the second you have more than 5 skills, you have an architecture problem: nothing knows about your brand, every skill imposes its own default tone, and outputs across skills don't compose.
You've probably tried: copy-pasting context into every prompt, maintaining a separate brand-context file you remember to attach, building a meta-skill that "loads brand." All work for a week, then drift back.
The fix isn't another skill. It's a shared compiled prompt every skill reads.
The architecture
brand-brain skill maintains 4 markdown files (positioning, voice, visual, story) as the source of truth. A compile step rolls them into one master-prompt.md. Every other skill reads it at the top of its instructions.
1
For whom · against what · what we promise · anti-promise
2
Tone axes · signature words · banned words · "without our name" test
3
Palette + typography + WHEN-TO-USE rules per scenario
4
Origin → core belief → transformation arc + through-line
Each pillar is its own file with versioned frontmatter (next-review: +90d). Update one pillar without touching the other three. Recompile on save. Every consuming skill picks up the change next invocation. No restart, no manual sync.
How the skills compose
BRAND BRAIN (Tier 1)
↓
master-prompt.md
↓
┌─────────────────┼──────────────────┐
↓ ↓ ↓
Pipeline (T2) Content (T2) Agency (T3)
┌─────────────┐ outreach website-audit
│ research │ ↓ cinematic-sites
│ ↓ │ repurpose personalized-banner
│ outreach │ ↓ site-caretaker
│ ↓ │ scripts/LinkedIn
│ funnel │
│ ↓ │
│ offer │
└─────────────┘
Lead-research outputs a structured brief. Outreach-writer reads it as input. Diagnostic-funnel reads outreach reply + brief. Offer-builder reads funnel output + brand brain → produces PDF. Four skills, one pipeline, voice consistent across all four because they all read the same compiled prompt.
All skills are plain markdown with YAML frontmatter. No custom runtime. Drop into ~/.claude/skills/, restart, done.
One-time purchase. No subscription. Lifetime updates by email. MIT license — fork it, modify it, use commercially.
Tier 1
Brand Foundation
$39~36 € · one-time
For: solo builder, one brand, foundational layer
Buy Tier 1⭐ Most popular
Tier 2
Growth Engine
$89~82 € · one-time
For: indie hacker / consultant / freelancer with active pipeline
Buy Tier 2Tier 3
Agency Suite
$199~184 € · one-time
For: small agency / studio / productized service
Buy Tier 3Production runs, not demos
Solo founder · personal brand
"Bootstrapped brand brain in 45 min. First weekly LinkedIn batch: 30 min instead of 2 hours. Voice consistency across 4 weeks of content: 100% (manually audited). Stopped explaining brand context to Claude every chat."
Consultant · €150 → €1.2K offer ladder
"lead-research → outreach-writer-pro → diagnostic-funnel → offer-builder. Average lead-to-offer time dropped from 2 days to 2 hours. First cohort of 20 cold DMs closed €8K deal. Pack paid back 90×."
Solo agency · 4 retainer clients
"Website audit as standalone product: €299 deliverable, 30-60 min of my time. Caretaker scheduled across all client sites. Zero regressions last quarter. Tier 3 paid back on a single client."
If you've already bought a Gumroad skill pack
Refund policy
Buy. Install. Run on a real project. If it doesn't pull its weight in your workflow within 7 days, reply to your purchase confirmation — refund within 24h, no friction.
After 7 days, refunds only for technical issues we can't resolve in 48h. In practice, never had to invoke this — Claude Code skills are environment-agnostic, they either work or they show a clear error.
Want to see actual skills before buying?
Mini brand-brain (Tier 1 in single-file form, ~200 lines) + Iskra case study PDF (45-min real bootstrap, full output) + Brand Brain checklist for printing. Read them, run them, then decide.
No upsell sequence. No "limited time offer expires in 17 minutes." One email, then nothing unless you reply.
Full functionality requires Claude Code (CLI / IDE / desktop) — that's where the skill auto-discovery and trigger system lives. With Claude Pro / web, you can paste SKILL.md content as Claude Project custom instructions for partial functionality (one skill at a time, manual). API integration is straightforward if you build your own dispatcher.
Outputs match the language you write to Claude in. SKILL.md (the skill specification) is in English for max compatibility. Brand voice rules can be in any language — write Polish voice rules, get Polish outputs in your brand voice.
No. Every skill works standalone with public web sources and manual input. lead-research-engine and outreach-writer-pro have optional integrations with enrichment APIs if you have them — but the standalone path is fully functional.
Yes — MIT license. Fork, modify, integrate into your own products, charge clients for work done with these skills. The only restriction: don't rebrand and resell as your own pack.
Handled by Stripe (via Gumroad). EU VAT collected per buyer's country. US sales tax per state. Receipt with invoice details available immediately after purchase.
Anthropic has a brand-voice marketplace, but it's enterprise-oriented (Notion / Box / Gong integrations) and lacks the connected-pipeline structure. This pack targets solo builders / freelancers / small agencies — different audience, different shape. The connected ecosystem (skills consuming each other's outputs) isn't replicable by isolated skills.
Lifetime, minimum 12 months guaranteed. Email with download link to new release. Your project-level brand brains stay untouched (they live in your workspace, not in the skills directory) — only skill logic updates.
Skills are markdown + frontmatter, not compiled code. After purchase you can drop the folder into your own private repo for version control. No GitHub for the source pack itself (yet) — keeps the distribution clean.
Tier 1: $39 · Tier 2: $89 · Tier 3: $199 — one-time, no subscription
Pays back the first productive week. 7-day refund, no questions.