Skip to content

CodeAlive-AI/ai-cofounder

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

AI Cofounder

AI Cofounder Banner

Your AI Chief of Staff for startup operations. 48 skills that turn any AI agent into an augmented CEO: strategic reviews, hypothesis tracking, GTM playbooks, founder sales, and more.

A compilation of frameworks and insights from Lenny Rachitsky's Podcast, Corey Haines' marketing playbooks, and Garry Tan (Y Combinator CEO). Get April Dunford, Elena Verna, Jen Abel, Bob Moesta, Marty Cagan, and dozens of other industry experts on your team, distilled into structured AI skills.

Built for founders who wear too many hats.

Who is this for

Technical founders — honestly, this is a must-have. You're good at building product but probably underinvesting in strategy, sales, and marketing. These 48 skills fill that gap: structured frameworks for everything from weekly CEO reviews to cold email outreach, without needing to hire a Chief of Staff or a sales coach. If you catch yourself hiding in code instead of talking to customers, this system will call you out.

CEOs and business leaders — a structured operating system for the hard decisions. Hypothesis tracking with WIP limits. Decision playbooks for pivots and hiring. Self-deception audits that flag when you're confusing activity with progress. A constraint-first approach that keeps you focused on what actually matters. Works whether you're pre-PMF or scaling.

Marketers — 24 GTM skills covering positioning, content strategy, SEO, cold email, social writing, growth loops, and launch playbooks. Each skill has frameworks from practitioners like April Dunford and Jen Abel. Use them as a second brain for marketing decisions, whether you're a one-person team or running a department.

Anyone building with AI agents — the skills are portable. Install them into Claude Code, Claude Cowork, Cursor, Codex, OpenClaw, or any of 40+ agents that support the skills format. They work standalone or as an integrated system. Claude Cowork users get these skills automatically (same ~/.claude/skills/ directory, no extra setup).

Install

Any agent (Claude Code, Claude Cowork, Cursor, Codex, etc.)

Install all skills at once:

npx skills add CodeAlive-AI/ai-cofounder -g

Or install individual skills:

npx skills add CodeAlive-AI/ai-cofounder@hypothesis-tracker -g
npx skills add CodeAlive-AI/ai-cofounder@decision-playbook -g
npx skills add CodeAlive-AI/ai-cofounder@founder-sales -g

No terminal? Just tell your agent: "Install skills from https://github.com/CodeAlive-AI/ai-cofounder following the README"

OpenClaw (full autonomous CEO OS)

The skills work in any agent, but OpenClaw is where the full system comes alive. The agent lives in your Telegram, WhatsApp, or Discord. It sends you insights on its own, runs daily recaps, tracks hypotheses across sessions, and nudges you when you're drifting from the constraint. You don't open it. It just works in the background.

git clone https://github.com/CodeAlive-AI/ai-cofounder.git
cd ai-cofounder/openclaw
./install.sh

Then customize for your company — see openclaw/CUSTOMIZATION.md. Full setup guide: openclaw/README.md.

What is this?

48 skills organized into a CEO operating system. Each skill is a structured workflow with frameworks from 100+ product leaders. They work standalone, but together they form a system:

  • Weekly CEO reviews with self-deception checks
  • Hypothesis tracking with WIP limits and evidence tiers
  • Decision playbooks for pivots, hiring, pricing, and more
  • Discovery debriefs that extract structured learnings from customer conversations
  • GTM skills covering positioning, sales, content, SEO, email, and launch
  • Proactive nudges via heartbeat — insights, coaching, and daily recaps (OpenClaw only)

The system is opinionated. It follows a constraint-first approach (Theory of Constraints), uses evidence tiers to separate signal from noise, and checks for self-deception patterns that founders usually don't notice on their own.

Repo structure

ai-cofounder/
├── skills/                  <- 48 portable skills (work in any agent)
│   ├── ceo-weekly-review/
│   ├── hypothesis-tracker/
│   ├── founder-sales/
│   └── ...
├── openclaw/                <- Full CEO OS for OpenClaw (Telegram/WhatsApp/Discord)
└── docs/                    <- Skill catalog, tool setup guides

Skills (48 total)

Category Count Examples
CEO Operating System 12 Weekly reviews, hypothesis tracking, decision playbooks, strategic analysis, anti-pattern audits
Go-To-Market 24 Positioning, founder sales, cold email, SEO, content strategy, pricing, user onboarding
Launch 2 Product Hunt launch playbook + 30-day timeline
Writing Quality 2 AI pattern removal (humanizer), prompt engineering
Tools (optional) 8 PostHog analytics, Exa research, web browser
Full skill catalog (click to expand)

CEO Operating System (12 skills)

Skill What it does
ceo-weekly-review Weekly review: dashboard, 10 questions, self-deception check, next week focus
hypothesis-tracker Add, update, rank, and kill hypotheses with WIP limits and evidence tiers
decision-playbook 9 quick-reference frameworks: pivot, hire, double down, kill feature, etc.
strategic-review Deep multi-phase analysis: premise challenge, alternatives, risk map, adversarial review
business-investigation Root cause analysis for business problems (why no signups, why churn)
discovery-debrief Structured debrief after customer conversations — extract learnings, update hypotheses
metrics-briefing CEO-level metrics through 9-category framework
anti-pattern-audit Monthly self-deception audit — 25 anti-patterns
idea-generator Generate business, marketing, and content ideas tied to your wedge
daily-recap End-of-day consolidation — summarize, promote durable facts to MEMORY.md
structured-log Write structured entries to daily memory for better recall
fpf-problem-solving First Principles Framework — rigorous decomposition for complex problems

Go-To-Market (24 skills)

Skill What it does
problem-definition Define and validate the problem you're solving
positioning-messaging Craft positioning, value props, and taglines
product-marketing-context Shared marketing context across skills
competitive-analysis Analyze competitors and find differentiation
conducting-user-interviews Plan and run effective customer interviews
measuring-product-market-fit Measure PMF with frameworks beyond Sean Ellis
writing-north-star-metrics Define your North Star metric
pricing-strategy Design pricing tiers and packaging
founder-sales Close first customers, build repeatable sales process
sales-qualification Qualify leads before investing time
enterprise-sales Navigate complex buying committees
product-led-sales Convert self-serve users to enterprise
content-strategy Plan what content to create and why
copywriting Write landing pages and web copy
social-writer Write LinkedIn and X posts that sound human
cold-email Write cold emails that get replies
email-sequence Design lifecycle and nurture email sequences
seo-audit SEO and GEO (generative engine optimization) audit
programmatic-seo Scale content with programmatic SEO
designing-growth-loops Design viral and content growth loops
user-onboarding Optimize user onboarding and activation
brand-storytelling Craft company narrative and pitch story
community-building Build and grow developer/user communities
launch-marketing Plan and execute product launches

Launch (2 skills)

Skill What it does
product-hunt-launch Product Hunt launch specs and tactics
ph-launch-timeline 30-day Product Hunt preparation plan

Writing Quality (2 skills)

Skill What it does
humanizer Remove AI-generated writing patterns
prompt-engineering Universal prompt engineering techniques

Tools (8 skills, optional)

These skills connect the agent to external services. 7 of them need mcporter, a CLI bridge that lets AI agents call MCP servers. Without it, the skills will install but won't be able to reach PostHog or Exa APIs.

# One-time setup
npm install -g mcporter
# Then configure your API keys — see docs/adding-tools.md
Skill What it does Requires
posthog-analytics Query PostHog data PostHog account + mcporter
exa-company-research Research companies via Exa Exa API key + mcporter
exa-people-research Research people and LinkedIn profiles Exa API key + mcporter
exa-personal-site Find personal sites and blogs Exa API key + mcporter
exa-x-search Search X/Twitter posts Exa API key + mcporter
exa-code-search Search code repositories Exa API key + mcporter
exa-lead-gen Deep search for lead generation Exa API key + mcporter
web-browser Browse web pages via agent-browser agent-browser CLI

The other 40 skills work out of the box with no external dependencies.

Key concepts

Evidence tiers

Not all signals are equal. The system uses 5 tiers to assess evidence quality:

  1. Tier 1: Costly-to-fake behavioral (paid pilots, expansion, renewals)
  2. Tier 2: Product evidence (session replays, drop-offs, support tickets)
  3. Tier 3: Commercial evidence (win/loss, objections, stalled deals)
  4. Tier 4: Market narrative (competitor moves, content performance)
  5. Tier 5: Low-power (praise, vanity traffic, investor excitement)

Hypothesis management

Hypotheses follow the format: "For [ICP] who [trigger], if we [change], then [metric] will move from X to Y by [date], because [evidence]."

WIP limits: max 3 active hypotheses, max 2 concurrent experiments. Relay race rule: if evidence is sufficient before the deadline — decide now.

Self-deception checks

25 anti-patterns organized in 5 blocks: general founder mistakes, ex-engineer mistakes, metric interpretation, pre-PMF growth, and AI-era pitfalls. The agent flags these during weekly reviews and monthly audits.

OpenClaw integration

The openclaw/ directory has the full workspace setup:

  • SOUL.md — AI Chief of Staff persona with constraint-first thinking
  • MEMORY.md — Structured company context (wedge, ICP, stage, constraint)
  • HEARTBEAT.md — Proactive nudge cadence (insights, coaching, daily recaps)
  • AGENTS.md — Skill routing and disambiguation rules
  • Memory system — Daily logs, weekly reviews, hypothesis tracker

See openclaw/README.md for setup instructions.

Works great with CodeAlive

CodeAlive is an AI context platform for codebases. It lets agents instantly get answers about how anything works in a repository (architecture, dependencies, cross-service patterns) without reading thousands of files. If your startup builds software, the codealive-context-engine skill gives the CEO OS direct access to your codebase for deeper technical context during strategic reviews and investigations.

Docs

What's inside

The CEO OS is built on five pillars:

  1. Lenny Rachitsky's Podcast — 14 GTM skills distilled from 100+ episodes with top product leaders. Frameworks for positioning (April Dunford), founder sales (Jen Abel), growth loops, PMF measurement, and more. Packaged by RefoundAI/lenny-skills (MIT).

  2. Corey Haines' Marketing Skills — 8 skills for copywriting, cold email, SEO, content strategy, email sequences, and programmatic SEO. Practical frameworks from a SaaS marketer. Source: coreyhaines31/marketingskills (MIT).

  3. Garry Tan's gstack — Inspired the strategic review modes (EXPAND / SELECTIVE / HOLD / REDUCE) and structured decision-making patterns.

  4. First Principles Framework (FPF) by Anatoly Levenchuk — Transdisciplinary reasoning architecture for rigorous decomposition of complex problems. Used in fpf-problem-solving skill.

  5. CEO Bible — A research document on the role of CEO in an early-stage startup, synthesized by GPT-5.4 Pro from 50+ sources (Goldratt's Theory of Constraints, Horowitz's Wartime CEO, Bezos's two-way doors, Jobs's subtraction, and others). Powers the CEO OS skills: weekly reviews, hypothesis tracking, anti-pattern audits, decision playbooks, and constraint-first thinking.

License

This project is MIT licensed.

Skills included from third-party sources retain their original licenses:

Source Skills License
RefoundAI/lenny-skills 14 MIT
coreyhaines31/marketingskills 8 MIT
blader/humanizer 1 MIT
inference-sh/skills 1 No license specified
ailev/FPF (spec only) 1 No license on spec; skill packaging is MIT
Original (this repo) 23 MIT

See SOURCE.md in each skill folder for detailed provenance.

Contributing

PRs with new skills and agent setups are welcome.

Before submitting, please check that your skill doesn't duplicate functionality already covered by an existing one — review the skill catalog and the descriptions in skills/*/SKILL.md. If there's overlap, consider extending the existing skill instead.

What we're looking for:

  • New skills that fill gaps in the CEO OS (finance, hiring, fundraising, legal, etc.)
  • Agent setups for other platforms (agents/claude-code/, agents/codex/, etc.)
  • Improvements to existing skills — better frameworks, more examples, clearer instructions
  • Bug fixes — broken cross-references, outdated advice, formatting issues