Run three linked workshops inside your AI coding agent. They build your offer, lead plan and customer copy. Separate reviewers attack the maths, message and plan while each weak spot is still cheap to fix.
“AI automation for small businesses.”
A plausible idea. Too vague to trust.
When two agents flag the same issue, the phase does not pass.
“A five-day enquiry follow-up system for independent garages, installed with your existing inbox.”
The workshop fixes the decision, then runs the critics again.
Stronger does not mean proven. Real customers still decide.
Most AI is helpful by default. It fills gaps, smooths rough edges and tells you the idea has potential. That is useful until the missing objection is the one that kills the sale.
“This is a compelling idea with strong market potential.”
“Your buyer is vague, the promise is unmeasurable and the delivery model will not scale. Fix those before pricing it.”
Each phase has a gate. Research and drafting happen first. The critics review independently. Critical issues go back through the rig.
Your agent asks the hard questions, researches missing facts and drafts the phase output.
A sceptical marketer, business strategist and dynamic customer personas review it in parallel.
Shared concerns become must-fix items. You revise, accept a trade-off or stop.
Approved decisions are written to files so the work survives restarts and hands off cleanly.
Use one workshop on its own or run the full chain from offer to acquisition to final copy.
Market choice, pricing, value equation, delivery, bonuses, scarcity, guarantees and naming. Every phase gets reviewed before it passes.
Inspect the workshopresearch.mdoffer.mdoffer-summary.htmlworkshop-progress.htmlLead magnets, Core Four channel selection, outreach scripts, ads, content, referrals and a Rule of 100 plan. Customer personas receive the outreach as prospects.
Inspect the workshopleads-blueprint.mdoutreach-scripts.mdtracking-dashboard.mdCustomer-facing copy goes through plain-language rules, de-AI checks and a reader panel when the stakes are high. The offer and lead workshops call it automatically.
Inspect the copy gateEach approved phase writes to plain files in your project. Restart the session, switch agents or hand the work to the next workshop without rebuilding the context from memory.
Open source. MIT licensed. Pick the install that matches your harness, then ask for the business problem in plain language.
/plugin marketplace add George-RD/growth-arsenal
/plugin install growth-arsenal@growth-arsenalgit clone https://github.com/George-RD/growth-arsenal
cd growth-arsenal
./install.sh codex # claude | omp | opencode | agents“I run a small bookkeeping firm. Help me build an offer I can charge five times more for.”
No. The workshops are phase-gated, write persistent files and run independent review agents at each checkpoint. A normal chat can imitate the tone, but it skips the structure and convergence rules.
No. It catches obvious gaps before customers see them. Dynamic personas are a pressure test, not proof of demand. Real buyers still decide whether the offer works.
No. The workshops apply the relevant questions and formulas one phase at a time. The repository is an independent open-source implementation inspired by $100M Offers and $100M Leads.
Yes. Each workshop runs on its own. The lead workshop can audit an existing offer, and the copy gate can review any customer-facing text.