YC New Way

Apply Y Combinator’s “New Way to Build a Startup” playbook to ship fast, validate with real distribution, and iterate using data. Use when the user asks to f...

Security Scan

This is an instruction-only skill that faithfully encodes the YC “New Way” playbook and requests no credentials, binaries, or installs — its requirements match its stated purpose.

Purpose & Capability

Name and description (YC New Way playbook → ship, distribute, learn) align with the included templates and step-by-step instructions. The skill does not request unrelated capabilities (no cloud creds, no unusual binaries).

Instruction Scope

SKILL.md and the reference templates stay within the playbook: clarifying hypothesis, picking a wedge, choosing distribution, minimal MVP, metrics, and weekly loop. There are no instructions to read local files, access environment variables, or call hidden endpoints. Guardrails explicitly discourage bypassing logins/CAPTCHAs and request evidence-first sourcing.

Install Mechanism

No install spec and no code files — instruction-only skill. Nothing is written to disk and no external archives are downloaded.

Credentials

The skill requires no environment variables, credentials, or config paths. All required inputs are user-provided planning details and publicly-citable evidence — proportional to the stated purpose.

Persistence & Privilege

always:false and model invocation is enabled (default). The skill does not request permanent presence or modify other skills or system settings.

Assessment

This skill is internally consistent and low-risk because it is instruction-only and asks for no secrets or installs. Before installing, consider: (1) limit the agent's network permissions if you don't want it performing live web scraping or publishing on your behalf; (2) follow the skill's own guardrail to avoid scraping behind logins or bypassing CAPTCHAs and respect Terms of Service; (3) when you track analytics or paste dashboard URLs, prefer trusted analytics providers and avoid sharing sensitive user data; and (4) review any outputs before using them as copy or publishing to ensure evidence links are accurate and no unintended PII is exposed.

YC “New Way” (ship → distribute → learn)

Goal

Turn an idea into a running distribution loop fast (days, not weeks), using the smallest product that can measure real demand.

Workflow (default)

0) Clarify the bet (2 minutes)

Fill this sentence:

• For [who] with [pain], we offer [promise] via [channel], and we’ll know it works when [metric] hits [threshold].

If missing, use references/experiment-brief.md.

1) Pick a narrow wedge

Prefer wedges that are:

• High intent (people already searching/buying)
• Clear objects (SKUs, listings, templates)
• Easy to prove with evidence (links, screenshots, timestamps)

Output:

• 1 primary wedge + 1 backup wedge
• 3 user stories

2) Choose one distribution channel first

Pick exactly one:

• SEO (content library + internal links)
• Newsletter (weekly digest)
• Community (Reddit/Discord/forums) where posting is allowed

Do not build more channels until one shows signal.

3) Ship the smallest MVP that measures

Default MVPs:

• SEO site: 10 pages (2 money pages + 8 support pages)
• Newsletter: 1 issue + signup + archive
• Tool page: 1 calculator/checklist with email capture

Include:

• Clear CTA
• Analytics
• Evidence links (avoid “AI says”)

Use references/landing-page-checklist.md and references/seo-10-page-plan.md.

4) Instrumentation & decision rules

Track only the fewest metrics:

• Traffic → CTA CTR → Outbound clicks → (optional) purchases
• Time to publish / pages per week (throughput)

Define stop/iterate rules:

• If CTR is low → fix promise/copy
• If CTR is high but no conversions → fix offer/price/trust
• If no traffic after 7 days → distribution problem (not product)

Use references/weekly-loop.md.

5) Weekly iteration loop

Every week:

1. Double down on 1 thing that worked
2. Kill 1 thing that didn’t
3. Add 1 new experiment

Keep a changelog; avoid rebuilding.

Output templates (use these verbatim)

• Experiment brief: references/experiment-brief.md
• Landing page checklist: references/landing-page-checklist.md
• SEO 10-page plan: references/seo-10-page-plan.md
• Weekly loop: references/weekly-loop.md

Guardrails

• Distribution before system: don’t over-automate until there’s signal.
• Evidence-first: recommendations must cite sources.
• Narrow first: one wedge, one channel, one metric.
• Legal/ToS: prefer public sources / allowed APIs; avoid bypassing logins or CAPTCHAs.