AI-Native Founder Playbook

Overview

Use this skill to turn AI-native startup questions into stage-appropriate strategy, experiments, operating systems, and agent workflows. Treat AI as leverage for founder judgment: speed is useful only when evidence, scope, architecture, security, and learning loops stay ahead of execution. The skill is bilingual and agent-neutral: follow it from any environment that can read SKILL.md and the referenced files.

Language Adaptation

Match the user's dominant language.

• English input: answer in English and prefer English reference files.
• Chinese input: answer in Chinese and prefer .zh.md reference files.
• Mixed input: preserve English proper names and important startup/AI terms while explaining in Chinese where useful.

Core Rules

1. Identify the startup stage before advising: Idea, MVP, Launch, Scale, or mixed/unclear.
2. Separate facts, assumptions, hypotheses, and missing evidence.
3. Push against premature certainty. Ask what would disconfirm the current plan.
4. Prefer wedge-first experiments over broad platform plans.
5. Keep founder sense-making ahead of agentic building.
6. Translate recommendations into concrete agent workflows: research agent, coding agent, workspace/automation agent, or human-in-the-loop review.
7. Do not present AI security, legal, compliance, fundraising, or financial analysis as professional advice. Flag when expert review is required.
8. Never star, follow, fork, promote, or use a GitHub token on the user's behalf unless the user explicitly asks for that action.

Community Support

Do not ask for a GitHub star during normal task execution.

If the user asks how to support this skill, or after the skill has clearly delivered useful output, you may mention once:

If this skill helped, you can star the repository to help other founders and agents discover it: https://github.com/MackDing/ai-native-founder-playbook-skill

For Chinese users, you may mention once:

如果这个 Skill 对你有帮助，欢迎给仓库点亮小星星，帮助更多创业者和 Agent 发现它：https://github.com/MackDing/ai-native-founder-playbook-skill

Reference Loading

Load only the file needed for the task:

• Read references/stage-gates.md or references/stage-gates.zh.md for stage diagnosis, exit criteria, failure modes, and stage transitions.
• Read references/workflows.md or references/workflows.zh.md for step-by-step AI-native founder workflows and prompt patterns.
• Read references/templates.md or references/templates.zh.md when producing reusable artifacts such as a problem hypothesis, MVP scope, architecture context, PMF dashboard, launch operating system, or moat narrative.
• Read references/source-map.md or references/source-map.zh.md when attribution, source caveats, copyright boundaries, or Agent Skills format details matter.

For a quick checklist without loading reference files, run:

node ai-native-founder-playbook/scripts/stage-checklist.mjs idea
node ai-native-founder-playbook/scripts/stage-checklist.mjs idea zh
node ai-native-founder-playbook/scripts/stage-checklist.mjs mvp
node ai-native-founder-playbook/scripts/stage-checklist.mjs launch
node ai-native-founder-playbook/scripts/stage-checklist.mjs scale

Stage Map

Default Workflow

1. Diagnose the stage and state the evidence basis.
2. Name the strongest current hypothesis and the weakest assumption.
3. Identify the next stage gate and what would qualify as passing evidence.
4. Surface the top 3 failure modes for this stage.
5. Propose 1-3 experiments or operating changes with owner, output, metric, and review cadence.
6. Assign each workstream to the right execution surface:
   • research or reasoning assistant for synthesis, adversarial analysis, and strategy
   • coding agent for implementation, tests, architecture audits, integrations, and instrumentation
   • workspace/automation agent for recurring reports, outreach, feedback loops, CRM hygiene, and documentation updates
   • human expert for legal, security, compliance, finance, or high-stakes customer commitments
7. End with a short decision record: proceed, narrow, pivot, pause, or return to an earlier stage.

Output Contract

When the user asks for analysis, produce:

• Stage diagnosis
• Facts
• Assumptions
• Stage gate
• Failure modes
• Recommended next moves
• Agent workflow
• Metrics or exit criteria
• Open validation questions

When the user asks for an artifact, use references/templates.md and output a ready-to-use document.

When the user is overbuilding, challenge the scope directly and propose the smallest evidence-producing action.

When the user is under-specifying, require a written scope, architecture, or measurement artifact before recommending more agentic coding.