yc-advisor

提供创业全周期决策支持,覆盖从构思验证、联合创始人选择、产品开发、融资策略到团队建设与规模化增长等关键环节,整合YC精选的443份深度资源,包括保罗·格雷厄姆等专家的原创文章、成功创业者访谈及系统化课程内容。

快捷安装

在终端运行此命令,即可一键安装该 Skill 到您的 Claude 中

npx skills add Agent-3-7/agent37-skills-collection --skill "yc-advisor"

YC Advisor

Overview

This skill provides access to Y Combinator’s comprehensive library of 443 startup resources - essays, podcast transcripts, and video transcripts from YC partners, successful founders, and industry experts.

How to Use This Skill (Tiered Retrieval)

Key Principle: Use quick-index for discovery, but ALWAYS load full source content before answering.

Step 1: Understand Context (for broad questions)

For broad questions, clarify the user’s context:

  • Stage: Pre-idea | Idea | Building MVP | Launched | Scaling
  • Type: B2B | Consumer | Hardware | AI/ML | Marketplace
  • Role: Technical founder | Non-technical | Solo | With co-founder(s)

Step 2: Discovery

  1. Load references/quick-index.md to scan available resources (~500 lines, grouped by topic)
  2. Identify 3-5 most relevant resources based on the question (line counts help estimate size)
  3. Check references/learning-paths.md if user is on a founder journey
  4. Check references/frameworks/ for decision questions - use glob references/frameworks/*.md to list files, then read specific ones
  5. For deeper search, use grep on references/summaries.md (too large to load fully)

Step 3: Deep Dive

  1. Find files using the code - use glob pattern references/{CODE}-*.md
    • Example: For code DZ, use glob references/DZ-*.md to find the file
    • WARNING: NEVER read index.yaml - it exceeds token limits (64K tokens)
  2. Load the FULL content of top 2-3 resources
  3. Read completely - do not skim
  4. Extract key insights, quotes, and actionable advice

Step 4: Synthesize Answer

  1. Combine insights from multiple sources
  2. Quote directly from source material when valuable
  3. Always cite author and title for each point
  4. Acknowledge tradeoffs and contradictions between sources
  5. Never answer from summaries alone - always load full source content

Topic Categories

The library covers these main areas (use for initial filtering):

  • Getting Started: Should you start? Startup ideas, order of operations, student founders
  • Co-founders: Finding, relationships, equity splitting, technical vs non-technical
  • Product: MVP, product-market fit, design, building for users
  • Fundraising: Seed, Series A, investor pitching, SAFEs, term sheets
  • Growth & Metrics: Growth strategies, KPIs, conversion, retention
  • Customers & Sales: Talking to users, first customers, pricing, enterprise sales
  • Hiring & Team: First hires, engineering teams, equity, management
  • Culture & Leadership: Building culture, CEO evolution, board management
  • Common Mistakes: Startup killers, financial health, when to quit
  • Pivoting & Launching: Pivot strategies, launch timing, press
  • Scaling: Later stage advice, unicorn characteristics
  • Mindset: Resourcefulness, handling rejection, goal setting
  • AI Startups: AI opportunity, moats, vertical agents, vibe coding
  • Founder Interviews: Airbnb, Stripe, Coinbase, Reddit, Twitch, DoorDash
  • Specialized: Hardware, biotech, dev tools, crypto, location
  • Joining Startups: Choosing a startup, stages, equity
  • YC Application: Application tips, process, YC effect
  • Legal: Startup mechanics, terms, agreements

Usage Guidelines

For Complex Decisions

Questions like “Should I start my own startup or co-found with someone?”:

  1. Load quick-index.md to identify relevant resources
  2. Read 3-5 full source files covering different perspectives
  3. Synthesize across sources - look for consensus and contradictions
  4. Present balanced view acknowledging tradeoffs
  5. Cite specific authors and titles
  6. Ask clarifying questions about user’s specific situation

For Factual Questions

Questions like “What are the most common mistakes that kill startups?”:

  1. Use quick-index.md to find the most authoritative source
  2. Load and read the full source file
  3. Present comprehensively - don’t over-summarize
  4. Cite the source

For Learning Journeys

When users want to learn systematically:

  1. Check references/learning-paths.md for curated sequences
  2. Guide them through resources in order
  3. Summarize key takeaways at each step

Resources

references/quick-index.md (Primary Discovery)

Lightweight index (~500 lines) grouped by topic. Each entry shows:

  • Code, title, author, type, line count, founder stage
  • Use this first - small enough to load fully
  • Use glob pattern references/{CODE}-*.md to find files by code

Detailed summaries with content previews (~4300 lines). Too large to load fully.

  • Use grep to search for specific keywords
  • Provides more context than quick-index when needed

references/index.yaml (Maintenance Only - DO NOT READ)

Structured metadata for all resources. Too large for runtime use (64K tokens). Used only by maintenance scripts. For filename lookups, use quick-index.md instead.

references/learning-paths.md

Curated resource sequences for common founder journeys:

  • First-time founder path
  • AI startup path
  • Fundraising path
  • And more…

references/frameworks/ (Use glob to list, NOT Read)

Decision frameworks for common questions. Use glob references/frameworks/*.md to list files. Available frameworks:

  • should-i-start-a-startup.md
  • solo-vs-cofounder.md
  • bootstrap-vs-raise.md
  • when-to-pivot.md
  • when-to-quit.md
  • technical-cofounder-needed.md

references/*.md

The 443 full-content source files. Each follows this structure:

# [Title]

**Author:** [Author Name]
**Type:** [Essay|Podcast|Video]
**URL:** https://www.ycombinator.com/library/[CODE]-[slug]

---

[Full content - essays, transcripts]

File naming: [CODE]-[descriptive-name].md (e.g., 8z-how-to-get-startup-ideas.md)