Found with Claude: Skills para Fundadores
sinned/found-with-claude
13 skills estruturadas para validar e construir startups do zero até Demo Day
Ver no GitHubPack com 13 skills estruturadas em Claude para guiar fundadores do problema até Demo Day. Cada skill processa insumos brutos (notas, transcrições, dados de uso) e gera artefatos estruturados prontos para ação, encadeando-se automaticamente no pipeline de validação de startup.
Instalação
npx convez-ai sinned/found-with-claude1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Mapear e validar oportunidades de problema real
- Sintetizar descobertas de clientes de entrevistas brutas
- Definir estratégia de produto e scope de protótipo
- Estruturar narrativa para fundraising
- Guiar contratação e preparação para Demo Day
Veredito

Ferramenta prática e bem-estruturada para fundadores. O grande valor está no encadeamento lógico entre skills e na capacidade de processar input bruto em outputs concretos, economizando tempo de síntese. Limita-se ao contexto de startups em estágio pré-produto e requer disciplina do usuário para manter corpus organizado.
README
A tradução para o português deste README ainda está na fila.
found-with-claude
A Claude Skills pack for founders building from idea to Demo Day.
Most AI coding tools start at code. Founders don't.
Founders start with uncertainty — is this problem real? Who has it? What should we build first? What signals tell us it's working?
This repo gives founders 13 structured Claude Skills for that entire journey. Each skill is a repeatable process, not a chatbot conversation. Drop in your raw material — messy notes, interview transcripts, usage data — and get structured output you can act on the same day.
The skills
| Phase | Question | Skill |
|---|---|---|
| Find a problem | Is this pain real and underserved? | /founder-opportunity-map |
| Customer discovery | Who has it and what do they say? | /customer-discovery-synthesizer |
| Customer simulation | What would real users say about this idea? | /synthetic-customer-panel |
| Product strategy | What do we build and why? | /product-strategy-brief |
| Prototype planning | What exactly are we building? | /prototype-scope |
| Prototype review | Does the prototype prove the thesis? | /prototype-review |
| First users | Who are they and how do we reach them? | /first-users-outreach |
| PMF sprint | Are we seeing the signals that matter? | /pmf-signal-review |
| Growth | What lever do we pull next? | /growth-experiment-backlog |
| Hiring | Who is the right first hire? | /founder-hiring-scorecard |
| Fundraising | What is the story investors need to hear? | /fundraising-narrative |
| Demo Day | Can we land this in 2 minutes on stage? | /demo-day-script |
| Weekly | What did we learn? | /weekly-founder-review |
Quickstart
# 1. Clone
git clone https://github.com/sinned/found-with-claude
cd found-with-claude
# 2. Open Claude Code
claude
# 3. Set up your startup workspace
/new-startup
# 4. Start at the beginning — or wherever you are
/founder-opportunity-map
No setup. No API keys. No boilerplate. Just open Claude Code and invoke a skill.
How to use it
Input: bring your raw material
Every skill works from rough input. Drop your actual notes, transcripts, and data into the relevant corpus folder. No cleanup needed.
work/[startup]/opportunity/founder-context.md— who you are, what you knowwork/[startup]/opportunity/founder-notes.md— raw observationswork/[startup]/customers/customer-transcript-1.md— interview transcripts
Output: structured artifacts, not chat
Each skill produces a specific document: a memo, a brief, a scorecard, a script. Outputs are saved in the relevant corpus folder so they feed directly into the next skill.
Chain: each output is the next skill's input
/founder-opportunity-map → opportunity memo
/customer-discovery-synthesizer → synthesis of interviews
/synthetic-customer-panel → stress-test before building
/product-strategy-brief → one-page brief
/prototype-scope → what to build, exactly
/prototype-review → did it prove the thesis?
/first-users-outreach → how to find the first 10 users
/pmf-signal-review → are the signals real?
...
Iterate: run skills multiple times
Run /founder-opportunity-map before talking to customers. Run it again after.
The difference between runs is what you learned.
Work is organized by startup
work/
.current ← active startup (e.g., "startup-1")
startup-1/
opportunity/ ← problem space, market analysis, founder context
customers/ ← interview transcripts, synthesis, simulated panels
product/ ← strategy brief, prototype scope, build evidence
traction/ ← outreach, PMF signals, experiments
team/ ← hiring scorecards
pitch/ ← fundraising narrative, demo day script
ops/ ← weekly reviews, loop state
startup-2/
...
Organized by knowledge corpus — not by skill. Every skill reads from and writes to
the appropriate folder. Run /new-startup to create a workspace. work/.current tracks
which is active.
The knowledge store at knowledge/[startup]/ accumulates structured facts as skills run —
designed for future export to an external durable store.
Example workspaces (read these first)
startup-1 and startup-2 are worked examples, not real companies — they show what
the skills produce when run end-to-end. Browse them to see the output before running
anything yourself. Their customer interviews are illustrative sample data, not records
of real conversations (provenance is noted inside each example).
| Example | Idea | How far it goes |
|---|---|---|
startup-1 | Auto-generated weekly investor updates for founders | Full arc: opportunity → discovery → strategy → prototype → outreach → PMF → fundraising |
startup-2 | Reduce chemical inputs in agriculture (AgroGuard — a resistance-aware decision layer for crop advisers) | Opportunity → discovery (5 sample interviews) → synthesis → panel → strategy brief → prototype scope |
When you run /new-startup, it creates your own startup-N alongside these and switches
work/.current to it — the examples stay for reference.
The idea-to-Demo-Day journey
| Step | Focus | Skill |
|---|---|---|
| 1 | Find a problem | /founder-opportunity-map |
| 2 | Customer discovery | /customer-discovery-synthesizer |
| 3 | Customer simulation | /synthetic-customer-panel |
| 4 | Product strategy | /product-strategy-brief |
| 5 | Build prototype | /prototype-scope → prototype-with-claude |
| 6 | First users | /first-users-outreach |
| 7 | PMF sprint | /pmf-signal-review |
| 8 | Growth | /growth-experiment-backlog |
| 9 | Hiring | /founder-hiring-scorecard |
| 10 | Fundraising | /fundraising-narrative |
| 11 | Demo Day | /demo-day-script |
| Ongoing | Review | /weekly-founder-review |
Where gstack fits
gstack is a global Claude Code skills pack for the engineering sprint: plan, build, review, QA, ship.
The handoff: /prototype-scope (here) defines what to build. gstack's
/office-hours and /autoplan plan it. /prototype-review (here) evaluates the result.
See docs/gstack-integration.md for details.
Where prototype-with-claude fits
The sibling repo prototype-with-claude
is the dedicated build environment for the prototyping step.
/prototype-scope(here) → defines what to build- Open
https://github.com/sinned/prototype-with-claude→ build and eval the AI behavior /prototype-review(here) → evaluate the result against the original strategy
Once you know the process: /loop
After you've run through the journey once and understand how the skills connect,
/loop can orchestrate the whole thing autonomously.
It runs every skill it can, and stops with exact instructions when it needs
real-world founder input — customer interviews, the prototype build, outreach,
experiment results. Drop the input, run /loop again, it resumes.
/loop # runs what it can, stops when it needs you
Loop state is saved in work/[startup]/loop-state.md between runs.
For Anthropic reviewers
Where to look:
CLAUDE.md— project instructions for Claude Code.claude/skills/— all 15 skills (13 journey skills +/new-startup+/loop)work/startup-1/andwork/startup-2/— two worked examples with sample transcripts and outputs (see "Example workspaces" above)docs/founder-journey-map.md— how the skills connect
What to run:
/new-startup
/founder-opportunity-map
/customer-discovery-synthesizer
/weekly-founder-review
What makes this different:
- Skills cover the full company-building loop, not just the coding phase
- Each skill is opinionated and structured — specific input, specific output
- Skills chain: each output is the next skill's input
- Work is organized per-startup so multiple ideas can coexist
The broader "Claude for…" pattern
prototype-with-claude→ Claude for AI feature prototypingfound-with-claude→ Claude for company building, idea to Demo Day- (future)
investing-with-claude→ Claude for sourcing, diligence, portfolio ops
Each is a self-contained Claude Skills pack for a professional context — turning Claude Code from a coding assistant into a workflow accelerator for the whole job.