founder-playbook

name: founder-playbook version: "1.0.0" description: | Diagnose where an AI-native startup is in its lifecycle (Idea → MVP → Launch → Scale) and route to the right stage discipline. Provides the four-stage map, per-stage goals/exit-criteria/failure-modes, and a vendor-neutral product matrix (conversational-research / agentic-coding / workflow-automation, with Claude Chat/Code/Cowork as reference). Use when a founder asks "where am I", "what should I focus on now", "am I ready to move to the next stage", "how do AI-native startups work", or wants an overview of the whole journey. For deep work inside a single stage, this skill hands off to founder-stage-idea / -mvp / -launch / -scale. allowed-tools: Read, Write, Edit, Glob, Grep, WebFetch, WebSearch

Founder Playbook — AI-Native Startup Lifecycle (router)

A stage-gated coach for building an AI-native startup. This skill locates a founder on the lifecycle and routes to the matching stage discipline. It encodes the process and judgment from Anthropic's The Founder's Playbook: Building an AI-Native Startup (see Attribution), generalized to be vendor-neutral.

The core reframe

In an AI-native startup the founder shifts from individual contributor to orchestrator of agents — specialized AI assistants that research, write and run code, and automate operations. AI has removed the three classic gates between stages: capital, headcount, and technical skill. The scarce resource is no longer building — it's judgment about what to build and whether it's working.

When to use this skill

• "I have an idea — where do I start?" / "where am I in the journey?"
• "Am I ready to move from MVP to Launch?" (exit-gate check)
• "What should I focus on / what's the biggest risk right now?"
• You want the whole-lifecycle overview or the tool-per-stage matrix.

When NOT to use it

• You already know your stage and want to do the work → go straight to the stage skill (founder-stage-idea | -mvp | -launch | -scale).
• You need legal/financial/medical/security sign-off → this is judgment support, not a substitute for a qualified human reviewer. Keep a human in the loop for nuanced calls (interpreting user feedback, security on auth/secrets/data, compliance).
• It's not a fundraising deck generator or a guarantee of product-market fit.

The four-stage map

Stage diagnosis flow

Ask the founder these, in order — stop at the first "no":

1. Do you have qualitative evidence (from real conversations, not surveys) that a specific group has this problem and that your solution addresses it?
   • No → Idea stage → load founder-stage-idea.
2. Do real users return to / pay for / refer the product (evidence of PMF, e.g. Sean Ellis ≥40% "very disappointed"; effort starts to pull not push)?
   • No → MVP stage → load founder-stage-mvp.
3. Is growth repeatable through known channels (CAC·LTV·payback you can defend) AND do operations run without you personally in every loop?
   • No → Launch stage → load founder-stage-launch.
4. Are you building systematic growth + a defensible moat, with the company sustainable even when you step back from day-to-day?
   • No (you cleared 1–3 but aren't there yet) → you're early in Scale → load founder-stage-scale.
   • Yes → you're maturing through Scale toward the threshold exit (profitability / IPO / acquisition); founder-stage-scale still applies — re-check its exit gate.

A founder can be mid-stage. If unsure between two, pick the earlier one — the earlier discipline is cheaper to revisit than the failure mode it prevents.

Product matrix (vendor-neutral)

Three capability classes power the lean startup; see references/product-matrix.md for the per-stage mapping. Claude (Chat / Code / Cowork) is the canonical reference implementation; the framework holds for any AAIF-compatible toolset.

Cross-stage anti-patterns (watch at every stage)

• Speed as the only variable — AI removes natural bottlenecks, so velocity is guaranteed; judgment is the scarce input. Easy prototyping inflates the risk of building something nobody needs.
• Skipping persistent context — without written specs/architecture/decisions (e.g., a CLAUDE.md/context file), each session re-derives foundations and drifts.
• Treating early energy as proof — launch spikes from friends, investor networks, or a front-page post don't predict week-6/week-12 retention.
• Vendor lock-in — keep the capability classes provider-agnostic so a single vendor outage or pricing change can't paralyze the company.
• Founder as permanent bottleneck — the job is to design the systems that do the work, not to keep doing all the work.

How to drive the family

1. Run the diagnosis flow → name the stage.
2. Load the matching stage skill for goal + exit-criteria checklist + failure modes + ready-to-use exercise prompts.
3. Re-check the exit gate before advancing; advancing early is its own failure mode.
4. For an interactive coach persona (delegatable via the Task tool), use the founder-coach agent. To invoke this router explicitly, run /founder-playbook.

Attribution

Framework adapted (process & methodology, original prose) from Anthropic, The Founder's Playbook: Building an AI-Native Startup (2026-05-14) — https://claude.com/blog/the-founders-playbook. No text is reproduced verbatim; this skill encodes the framework for agent use under the repository's MIT license.