convez.ai
SkillProdutividadeIntermediário
sinned

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 GitHub
InicialPluginScore · de 0 a 100

Pack 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-claude

1 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

Estrelas
0
Forks
0
Acompanhando
0
Saúde
28%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

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

PhaseQuestionSkill
Find a problemIs this pain real and underserved?/founder-opportunity-map
Customer discoveryWho has it and what do they say?/customer-discovery-synthesizer
Customer simulationWhat would real users say about this idea?/synthetic-customer-panel
Product strategyWhat do we build and why?/product-strategy-brief
Prototype planningWhat exactly are we building?/prototype-scope
Prototype reviewDoes the prototype prove the thesis?/prototype-review
First usersWho are they and how do we reach them?/first-users-outreach
PMF sprintAre we seeing the signals that matter?/pmf-signal-review
GrowthWhat lever do we pull next?/growth-experiment-backlog
HiringWho is the right first hire?/founder-hiring-scorecard
FundraisingWhat is the story investors need to hear?/fundraising-narrative
Demo DayCan we land this in 2 minutes on stage?/demo-day-script
WeeklyWhat 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 know
  • work/[startup]/opportunity/founder-notes.md — raw observations
  • work/[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).

ExampleIdeaHow far it goes
startup-1Auto-generated weekly investor updates for foundersFull arc: opportunity → discovery → strategy → prototype → outreach → PMF → fundraising
startup-2Reduce 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

StepFocusSkill
1Find a problem/founder-opportunity-map
2Customer discovery/customer-discovery-synthesizer
3Customer simulation/synthetic-customer-panel
4Product strategy/product-strategy-brief
5Build prototype/prototype-scope → prototype-with-claude
6First users/first-users-outreach
7PMF sprint/pmf-signal-review
8Growth/growth-experiment-backlog
9Hiring/founder-hiring-scorecard
10Fundraising/fundraising-narrative
11Demo Day/demo-day-script
OngoingReview/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.

  1. /prototype-scope (here) → defines what to build
  2. Open https://github.com/sinned/prototype-with-claude → build and eval the AI behavior
  3. /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/ and work/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 prototyping
  • found-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.

Ver o original no GitHub

Bom para

Relacionados